Overmodulation protection system for motor inverters based on SiC device junction temperature estimation

By using a SiC device protection system that monitors and predicts current and junction temperature in real time, the problem of inconsistent protection judgments during overmodulation switching is solved, the risk of false release is reduced, and the safety and reliability of motor inverters are improved.

CN123121422APending Publication Date: 2026-11-17JIANGSU AEROSPACE POWER ELECTRIC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611618870.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies, due to control lag and electrothermal coupling during overmodulation switching, result in inconsistencies between protection judgment criteria and actual execution status, posing a risk of current deviation or junction temperature exceeding limits.

Method used

The motor inverter overmodulation protection system based on SiC device junction temperature estimation uses sampling circuits and controllers to monitor phase current, bus voltage and module temperature in real time, predict the current and junction temperature state during overmodulation, allow overmodulation to exit only when protection constraints are met, and re-evaluate protection measures when state changes or parameters are inconsistent.

Benefits of technology

It reduces the risk of false releases caused by state changes and parameter changes during sampling to loading, ensures consistency of current and junction temperature predictions, reduces the risk of current deviation and junction temperature exceeding limits during overmodulation switching, and improves the reliability of the protection system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123121422A_ABST
    Figure CN123121422A_ABST
Patent Text Reader

Abstract

This invention discloses an overmodulation protection system for motor inverters based on SiC device junction temperature estimation, comprising a three-phase inverter bridge, a gate drive circuit, a sampling circuit, and a controller. When a junction temperature warning is triggered, the controller establishes two trajectories: one maintaining overmodulation permission and reducing torque, and the other lowering the modulation limit and reducing torque. A pulse segment that cannot be changed by the current update before the target loading time is used as a common pre-load pulse segment. The current and junction temperature at the end of this segment are predicted, and a candidate pulse sequence is generated. After dead-time and minimum pulse width processing, the current and junction temperature are predicted based on the same sequence. The corresponding trajectory is allowed to execute when protection constraints are met. The exit permission is bound to the verified first-cycle pulse parameters and the target loading time. The permission is revoked when the parameters change or the loading time is missed, thus reducing the risk of misjudgment caused by inconsistencies between the prediction basis and actual execution.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power electronic conversion and motor drive protection technology, specifically to an overmodulation protection system for motor inverters based on SiC device junction temperature estimation. Background Technology

[0002] Motor inverters using SiC devices supply the required voltage to the motor by adjusting the switching state. In operating ranges where bus voltage is limited, overmodulation can be used to extend the output voltage range. Device junction temperature estimation provides a basis for limiting the inverter's thermal load and determining protection actions.

[0003] Existing inverter thermal protection typically utilizes operating information such as phase current, bus voltage, and module temperature, combined with the relationship between device losses and thermal response, to estimate junction temperature, and adjusts torque command, modulation range, or output state based on the estimation results. For switching between overmodulation and linear modulation, existing control methods determine switching conditions based on modulation ratio, switching state, or current response to suppress output disturbances caused by mode switching. Under conditions where load changes are gradual, control parameters take effect at predetermined times, and thermal state changes can be reflected in the estimation results, this approach can provide a feasible basis for output limiting and modulation switching.

[0004] However, during overmodulation operation accompanied by torque adjustment, sampling, computation, and pulse parameter activation occur at different time points. The device still undergoes conduction and commutation processes between these points, and its current state and internal thermal state may continue to change. The fact that the junction temperature at the sampling moment has not reached the protection limit does not mean that the control command will still have the same temperature margin when it takes effect. When parameter updates are delayed or the parameters to be executed change, the previously determined operating state may deviate from the actual operating state. On the other hand, reducing the torque command or lowering the modulation range will simultaneously affect the current component, conduction interval, and commutation frequency. The direction of change in conduction loss and switching loss may differ, making it difficult to determine whether the device's thermal load has decreased based solely on a reduction in a single command value. Dead zones and narrow pulse processing may also alter the effective voltage and commutation process at the switching boundary, causing discrepancies between the current and temperature rise judgments based on the ideal modulation state and the actual output. For processing methods that use current junction temperature compliance or command reduction as the primary basis for release, the aforementioned timing offsets and electrothermal coupling may cause the modulation switching that was originally deemed executable to actually result in current deviation or junction temperature exceeding the limit. Summary of the Invention

[0005] The purpose of this invention is to provide an overmodulation protection system for motor inverters based on SiC device junction temperature estimation, in order to solve the technical problem in the prior art that the protection judgment basis is inconsistent with the actual execution state due to the control lag and electrothermal coupling during overmodulation switching.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an overmodulation protection system for motor inverters based on SiC device junction temperature estimation, comprising a three-phase inverter bridge with SiC switching devices, a gate drive circuit, a sampling circuit, and a controller, wherein the controller controls the SiC switching devices through the gate drive circuit;

[0007] The sampling circuit is connected to the controller to collect phase current, bus voltage, rotor position and module temperature;

[0008] The controller updates the current junction temperature state of each SiC switch based on the sampling results and the loaded pulse sequence. When overmodulation occurs and the junction temperature of any SiC switch reaches the warning threshold, a first trajectory that maintains overmodulation permission and reduces torque, and a second trajectory that reduces the modulation limit and reduces torque are established. The pulse segment that cannot be changed by this update before the target loading time is used as a pre-pulse segment shared by the two trajectories. The current state and junction temperature state at the end of the pre-pulse segment are predicted. Candidate pulse sequences connected to the pre-pulse segment are generated with the prediction results as the initial state. After dead-time and minimum pulse width processing based on the end switching state of the pre-pulse segment, the applied voltage and device loss are calculated using the same candidate pulse sequence to obtain the predicted current and predicted junction temperature. The corresponding trajectory is allowed to be executed only when the predicted current and predicted junction temperature meet the protection constraints. The exit permission of the second trajectory is bound to the first cycle pulse parameter of the candidate pulse sequence that has passed the verification and the target loading time. The exit permission is revoked when the loading parameters change or the target loading time is missed.

[0009] Furthermore, the controller includes a pulse width modulation timer; the target loading time corresponds to an update event in which the pulse width modulation timer synchronizes the three-phase pulse parameters; the preceding pulse segment is a pulse segment from the sampling reference time to the target loading time that has been determined by the loaded pulse sequence and cannot be changed by this modulation update.

[0010] Furthermore, the controller recursively obtains the current state and junction temperature state at the end of the pre-pulse segment based on the conduction interval and commutation event in the pre-pulse segment, combined with the phase current and the bus voltage, and records the conduction state, the most recent commutation time and the remaining dead time of each SiC switching device at the end of the pre-pulse segment as a common boundary condition for generating the candidate pulse sequences corresponding to the two trajectories.

[0011] Furthermore, the controller checks the boundary commutation of the candidate pulse sequence according to the common boundary conditions, and adjusts only the candidate pulses from the target loading time so that the commutation between the complementary SiC switching devices meets the preset dead time and the pulse duration meets the preset minimum pulse width; after adjustment, the applied voltage and the device loss are recalculated; for candidate pulse sequences that require changes to the preceding pulse segment to meet the preset dead time or the preset minimum pulse width, their corresponding trajectory execution is prohibited.

[0012] Furthermore, the controller stores a device loss parameter table and transient thermal network parameters; the device loss parameter table includes conduction loss parameters and commutation energy parameters corresponding to current, bus voltage, temperature and gate drive conditions; the controller calculates the device loss of the executed period based on the loaded pulse sequence, updates the current junction temperature state in combination with the module temperature and the transient thermal network parameters, and then recursively obtains the corresponding predicted junction temperature based on the device loss of each candidate pulse sequence.

[0013] Furthermore, the controller includes a current regulator and a field weakening regulator; the controller uses the phase current, the rotor position, the integral state of the current regulator and the limiting state of the field weakening regulator as inputs for current prediction, performs current recursion based on the applied voltage corresponding to the candidate pulse sequence, and performs the same current limiting, field weakening regulation and anti-integral saturation processing as actual control during the recursion process.

[0014] Furthermore, the protection constraints include phase current limits, field weakening current limits, current tracking deviation limits, and device junction temperature limits. During the prediction period covering the preceding pulse segment and the subsequent candidate pulse sequence, the controller checks the current and junction temperature at each prediction moment and determines the temperature protection margin based on the junction temperature prediction error obtained from offline calibration within the corresponding operating range. Only when the phase current amplitude, field weakening current component, and tracking deviation relative to the corresponding current command of the predicted current satisfy the phase current limits, the field weakening current limits, and the current tracking deviation limits, respectively, and the predicted junction temperature of each SiC switching device, after being superimposed with the temperature protection margin, does not exceed the device junction temperature limit, is the corresponding trajectory deemed to satisfy the protection constraints.

[0015] Furthermore, the controller establishes an authorization record for the first cycle pulse parameters and the target loading time corresponding to the second trajectory; before triggering the pulse loading corresponding to the update event, the parameters to be loaded are checked against the authorization record, and loading is allowed only when the check is consistent and the target loading time is not missed, and the authorization record is invalidated after loading; if the check is inconsistent or the target loading time is missed, the exit action is not performed and a re-evaluation is performed.

[0016] Furthermore, when the second trajectory satisfies the protection constraint and the exit permission is valid, the controller executes the first control step of the second trajectory; when the second trajectory does not satisfy the protection constraint but the first trajectory does, the controller executes the first control step of the first trajectory, and regenerates and evaluates the two trajectories based on the sampling results after execution; when neither of the two trajectories satisfies the protection constraint, or when the cumulative execution time of the first trajectory since the current junction temperature warning is triggered reaches a preset waiting time limit, the controller prohibits the continued execution of the overmodulation exit process and outputs a fault handling request.

[0017] Furthermore, after executing the first control step of the first trajectory or the second trajectory, the controller updates the current junction temperature state according to the actual loaded pulse parameters and compares the re-acquired phase current with the predicted current at the same time. When the actual loaded pulse parameters are inconsistent with the verified pulse parameters, or the current comparison deviation exceeds the preset deviation limit, the subsequent overmodulation exit step is prohibited, and the two trajectories are re-evaluated based on the actual loaded pulse parameters and the re-acquired phase current. When the junction temperature reaches the preset emergency protection threshold or both trajectories fail to meet the protection constraints after re-evaluation, the fault handling request is output.

[0018] Beneficial effects

[0019] 1. This invention incorporates the pre-pulse segment that cannot be changed by this update into the state prediction of the two trajectories, and limits the exit permission to the verified first cycle pulse parameters and the target loading time, so that the protection judgment corresponds to the current and junction temperature state when the candidate action can start to be executed, which helps to reduce the risk of false release caused by state changes, parameter changes or loading delays during the sampling to loading period.

[0020] 2. This invention processes boundary commutation based on the end state of the preceding pulse segment, and calculates the applied voltage and device losses using the same candidate pulse sequence after processing, so that the current prediction and junction temperature prediction correspond to the conduction interval and commutation event. Combined with transient thermal network, field weakening regulation state and temperature protection margin, the changes in current and thermal load caused by torque reduction and adjustment of modulation range can be included in the protection constraints, reducing the risk of judging thermal load reduction based solely on command reduction.

[0021] 3. In addition, by sampling verification, permission failure handling and waiting time limit constraints after the first step is executed, the subsequent exit steps are stopped and re-evaluated or a fault handling request is output when the loading parameters are inconsistent, the current deviation exceeds the limit or the protection conditions are not met. This helps to limit the continued use of the original judgment after the operating status changes. Attached Figure Description

[0022] The invention will now be further described with reference to the accompanying drawings.

[0023] Figure 1 This is a schematic diagram showing the system composition and connection relationships of this system.

[0024] Figure 2 This is a flowchart of the overmodulation protection and two trajectory selection for this system.

[0025] Figure 3 This is a timing diagram showing the pre-pulse segment and the target loading time of this system.

[0026] Figure 4 This is a flowchart of the candidate pulse cross-boundary commutation processing for this system.

[0027] Figure 5 This is a schematic diagram of current and junction temperature prediction based on the same candidate pulse sequence in this system.

[0028] Figure 6 This is a flowchart of the system's exit permission and pre- and post-installation verification process.

[0029] Explanation of reference numerals in the attached figures: 100, three-phase inverter bridge; 200, gate drive circuit; 300, sampling circuit; 400, controller; 410, pulse width modulation timer; 420, state estimation function; 430, trajectory generation function; 440, constraint verification function; 450, permission management function; 500, motor; 600, DC bus. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] The following combination Figures 1 to 6 This describes one implementation method of the system. The accompanying diagrams illustrate component connections, information flow, and control sequence. Timing waveforms are used to illustrate the relationships between events and do not represent measured waveforms. Device parameters, protection limits, and error margins in the implementation should be determined based on the selected inverter, motor, and heat dissipation assembly. Calibrations or verifications not yet performed are not presented below as existing test results.

[0032] System composition and connection relationships, such as Figure 1As shown, the system includes a three-phase inverter bridge 100, a gate drive circuit 200, a sampling circuit 300, and a controller 400. A DC bus 600 is connected to the DC side of the three-phase inverter bridge 100, and the AC side of the three-phase inverter bridge 100 is connected to a motor 500. This embodiment uses a two-level inverter bridge with six SiC switching positions. The SiC switching devices in the upper and lower arms of each phase are driven by pulse signals output from the controller 400 via the gate drive circuit 200. The motor 500 is a three-phase permanent magnet synchronous motor, and the neutral point of the three-phase windings is not connected to the DC bus.

[0033] The sampling circuit 300 receives phase current, DC bus voltage, rotor position, and module temperature signals, and provides the controller 400 with information including the sampling reference time. Phase current can be measured by two-phase current sensors and the third phase reconstructed based on the three-phase current relationship, or the three-phase current can be measured directly. The rotor position is obtained by a rotary transformer or encoder, and the rotational speed is determined by continuous position sampling. The module temperature is obtained by a temperature sensing element located within the power module or at a designated substrate location. The module temperature measurement location and dynamic response are recorded as thermal model calibration conditions.

[0034] The controller 400 incorporates a pulse width modulation (PWM) timer 410 and implements state estimation 420, trajectory generation 430, constraint verification 440, and permission management 450 through programming. These functions can be implemented within the same microcontroller, eliminating the need for independent processors. The PWM timer 410 displays active parameters, parameters to be loaded, and update events that synchronize the three-phase parameters. The permission management function 450 controls whether the verified parameters are loaded at the specified update event.

[0035] The controller 400 stores motor parameters, component loss parameter tables, and transient thermal network parameters. Motor parameters include stator resistance, direct-axis inductance, quadrature-axis inductance, and permanent magnet flux linkage, determined based on motor parameter data or conventional identification results; when magnetic saturation needs to be considered, parameter tables within the corresponding operating range can be used. The current regulator and field weakener regulator use coordinate transformation, limiting, and anti-integral saturation rules consistent with actual motor control.

[0036] When the rotor position detection result is a mechanical angle, the controller 400 converts the electrical angle according to the number of motor pole pairs, and calculates the electrical angular velocity from the same electrical angle sequence for use in coordinate transformation and current prediction.

[0037] State initialization and operational information updates: After each valid sample, the state estimation function 420 updates the current state, regulator state, and internal state of the thermal network based on the actual loaded pulse parameters, counter operation records, and sampling results since the previous sample. The current junction temperature state includes the internal state of the thermal network and its corresponding junction temperature output; a single instantaneous junction temperature value cannot replace the history of accumulated heat. Two candidate trajectories are predicted using state replicas, whose internal variables are separated from the actual control state.

[0038] During normal continuous operation, the historical state of the thermal network is preserved. When the temperature is sufficiently homogenized in a cold state, the thermal nodes can be initialized based on the module temperature. During a hot restart, a reliable historical thermal state should be restored, or a conservative initial value determined by the operating range before shutdown should be used, and new exit actions should be postponed until the state is reliable. The readings of the module temperature sensing elements are not directly equivalent to the SiC chip junction temperature, nor are they directly used as the chip case temperature without calibration.

[0039] like Figure 2 As shown, the controller 400 enters protection evaluation when overmodulation operation occurs and the estimated junction temperature at any SiC switch position reaches the warning threshold. The warning threshold is lower than the emergency protection threshold, and the interval is determined by the junction temperature estimation error, normal update delay, and allowable temperature rise. Hardware short-circuit protection, drive faults, and the device's original emergency protections maintain higher priority and cannot be overridden by the system's exit permission.

[0040] The timing of the pre-pulse segment and the target loading time, such as Figure 3 As shown, the reference time for this sampling is denoted as... The target loading time that will make the candidate first cycle parameters take effect synchronously is denoted as... The controller 400 reads the count value, counting direction, activity period, and comparison parameters of the pulse width modulation timer 410, and, in conjunction with the already determined loading plan, reconstructs the pulse segment between two moments that cannot be changed by the current normal modulation update. The contents of the register to be loaded, which can still be overwritten by the current operation, are not automatically considered unchangeable simply because they have already been written to the register.

[0041] The target loading time corresponds to a specific counter update event, such as the zero-crossing event of a three-phase synchronous counter. The event sequence number is recorded simultaneously to avoid confusion between different cycles based solely on identical count values. When selecting this event, worst-case operation time, register write time, and hardware loading setup time should be allowed. If processing cannot be completed before this event, a subsequent event can be selected, but the extended pre-pulse segment must be redefined and the prediction re-performed; simply changing the event sequence number and using the original result is insufficient.

[0042] The pre-pulse segment starts from the sampling reference time and includes the conduction interval, commutation events, and dead time within it. The controller 400 first recursively calculates to the end of this segment to obtain the current, regulator, and thermal network status at the target loading time, which serves as the prediction starting point for the two trajectories. Simultaneously, it records the end-conductivity status of each switch position, the most recent commutation time, and the remaining dead time. If a hardware fault forces a change in the output, the original pre-pulse segment becomes invalid, and the controller switches to the corresponding fault handling procedure.

[0043] The calculation interval of the pre-pulse segment includes the sampling reference time but not the target loading time; the commutation that occurs exactly at the target loading time is included in the candidate sequence after boundary processing to avoid the same commutation event being counted repeatedly in the two segments before and after.

[0044] The formation of two candidate tracks involves two tracks: the first track maintains the current permissible overmodulation upper limit and reduces the torque command according to a preset rate of change; the second track reduces the torque command while gradually reducing the modulation upper limit until it reaches the linear modulation range. For regenerative operation, reducing torque refers to reducing the absolute value of the torque command. Maintaining overmodulation permissibility means not further lowering its permissible upper limit and does not require the actual waveform to be forcibly overmodulated in each subsequent cycle.

[0045] This embodiment can use the same initial torque derating rate for both tracks, so that the comparison mainly reflects the impact of modulation upper limit adjustment. The torque derating rate is set according to the motor's allowable current, the drive system's torque change requirements, and the current field weakening state; the modulation upper limit rate is set according to the achievable range of the existing modulator. It does not presuppose that the phase current will necessarily decrease after torque reduction, nor does it presuppose that the switching loss will necessarily decrease after exiting overmodulation.

[0046] The trajectory generation function 430 copies the integral state of the current regulator, the limiting state of the field weakener regulator, and the required historical variables from the common prediction starting point. In each prediction step, it first generates a current command, then performs the same current regulation, field weakening regulation, limiting, and anti-integral saturation processing as in actual control, forming a voltage command and candidate pulses. The prediction length covers at least the preceding pulse segment and one complete cycle after target loading; longer prediction intervals are determined based on temperature rise changes, the effective range of the model, and the calculation budget.

[0047] Cross-boundary commutation and minimum pulse width processing, such as Figure 4 As shown, the candidate pulse must be connected to the state at the end of the preceding pulse segment. If the complementary device is just turned off before the target loading time, the turn-on time of the candidate device should meet the remaining dead time requirement; if a device is already turned on at the end of the preceding pulse segment, when calculating its minimum pulse width, the on-time that has been continuous before the target loading time should be included in the complete pulse, rather than requiring the preceding and following segments to reach the minimum pulse width separately.

[0048] Boundary legality can be checked under the following time conditions:

[0049]

[0050] In the formula, Candidate opening time, The most recent turn-off time of the complementary device in the same bridge arm. Preset dead time; , These represent the start and end times of the same continuous conduction pulse. The minimum pulse width is preset. The dead time and minimum pulse width are determined based on the selected device, drive propagation delay and differences, timer resolution, and drive allowable pulse width, while retaining applicable manufacturing and temperature margins.

[0051] During processing, only the target loading time and subsequent candidate pulses are allowed to be modified. Methods such as delayed turn-on, changing the adjustable turn-off time, or canceling the adjustable narrow pulse can be used, but the processing result must be expressible by the actual counting mode and the comparison register. Candidate sequences that require modification of the preceding pulse segment, cause complementary conduction conflicts, or exceed the timer's expressive capacity are eliminated. After a pulse change, the original voltage and loss calculation results are no longer used; instead, they are recalculated based on the changed waveform.

[0052] Voltage and current prediction under the same pulse sequence, such as Figure 5 As shown, the state estimation function 420 and the constraint verification function 440 use the same candidate pulse sequence after boundary processing. For the three-phase bridge arm, the output node potential is determined according to each conduction interval and commutation event. The output potential during the dead time is determined by the current direction and the freewheeling path; the fact that both upper and lower transistors are turned off cannot be interpreted as the output voltage of that phase being zero. When the current is close to zero and the direction cannot be reliably determined, the possible freewheeling states can be calculated separately, and the more unfavorable constraint result can be adopted.

[0053] In this embodiment where the three-phase neutral point is not connected to the busbar, the average applied voltage of phase a within a calculation interval can be expressed as:

[0054]

[0055] In the formula, The average phase voltage of phase a. This refers to the bus voltage of the interval. , , This represents the proportion of time during which the three-phase output nodes are effectively connected to the upper bus within this interval. This is a correction for the phase voltage caused by factors such as the on-state voltage drop of the device. The above time ratio has taken into account boundary treatment and dead zone freewheeling paths, and is not equivalent to the unprocessed ideal duty cycle; the other two phases are obtained by cyclic replacement according to the phase sequence. The bus voltage can be maintained at the current sampled value within a short interval, or it can be extrapolated for a short time based on the obtained rate of change.

[0056] After transforming the three-phase voltage to a rotating coordinate system using the coordinate transformation for actual current control, the current can be recursively derived using the following discrete motor model:

[0057]

[0058]

[0059] In the formula, and These are the d-axis and q-axis currents, respectively. and These are the corresponding axis voltages. For stator resistance, and These are the d-axis and q-axis inductances, respectively. Electric angular velocity, It is a permanent magnet flux linkage. The duration of the nth calculation interval. The electric angular velocity is obtained by sampling the existing rotor position and processed according to the agreed hold or extrapolation rules within the short prediction interval; the sign and direction are consistent with the coordinate transformation used in actual motor control.

[0060] The above recursion is an optional discrete implementation. The calculation step size should be sufficiently small relative to the electrical time constant and electrical angular velocity. If necessary, it should be subdivided according to the conduction and commutation intervals, or a calibrated discrete motor model can be used. The pulse-trimmed voltage is used for the current recursion in this step, and the updated current is then used for subsequent control calculations and loss estimation, avoiding the use of future measured currents that are not yet available as input.

[0061] Device loss and junction temperature state recursion

[0062] For each switching position, the losses of its forward channel, reverse channel, and body diode conduction paths are accumulated separately, and the turn-on or turn-off energy is retrieved according to the actual commutation event in the candidate waveform. The inputs to the loss table include current, bus voltage, temperature, and gate drive conditions; the initial values ​​of the table can be obtained from the device datasheet, and should be matched with the actual drive voltage, gate resistance, and thermal assembly conditions during application. Recovery losses already included in the turn-on energy table are not accumulated again.

[0063] The average loss of the r-th switch position within the n-th calculation interval can be expressed as:

[0064]

[0065] In the formula, This represents the average loss over the interval, expressed in watts. The instantaneous conduction loss is determined based on the current, temperature, and conduction path at that time, and is taken as zero during non-conducting periods; and The start and end times of the interval; This is the set of commutation events at that switch position; The energy of the corresponding event is expressed in joules. The predicted temperature used for table lookup is obtained from the initial thermal state of this step and is updated with the thermal state in subsequent calculation intervals.

[0066] This embodiment employs a transient thermal network including the chip, substrate, and temperature detection position response, using the module temperature as a reference input after assembly calibration. The relationship between the discrete state equation and the junction temperature output can be expressed as:

[0067]

[0068]

[0069] In the formula, This represents the internal state vector of the thermal network. This is a vector composed of the average losses at each switch position. This is the module temperature reference input. This is an estimated vector of representative junction temperatures for each switch position; , , , and This corresponds to the heat dissipation network matrix. The matrix is ​​obtained by discretizing network parameters with positive thermal resistance and positive thermal capacity, and calibrated in conjunction with the actual thermal response of the installation; when thermal coupling within the same module needs to be considered, corresponding coupling terms are retained between the loss input and the thermal state.

[0070] For different calculation intervals, the same continuous thermal network can be discretized according to the corresponding duration, or a pre-generated interval parameter table can be used. A mismatched discrete matrix cannot be directly reused after changing the step size. When the module temperature sensing element exhibits dynamic hysteresis, this response is incorporated into the thermal network identification; a single temperature bias is not used to replace dynamic calibration. For modules consisting of multiple chips connected in parallel at a single switch position, this embodiment estimates the calibrated representative junction temperature and does not claim that the highest temperature of each parallel chip can be identified solely based on existing sampling data.

[0071] The protection constraint and trajectory selection, constraint verification function 440 checks the phase current amplitude, field weakening current component, current tracking deviation, and device junction temperature at each calculation moment within the prediction period covered by the preceding pulse segment and candidate subsequent sequences. Phase current limits are determined based on the power device, motor, and current detection range; field weakening current limits are determined based on the motor's allowable operating range; tracking deviation limits correspond to the actual allowable deviation of current control. Each limit is determined by specific equipment conditions and is not replaced by fixed, universal values.

[0072] Junction temperature protection constraints can be written as:

[0073]

[0074] In the formula, Let r be the estimated junction temperature at the nth prediction time for the r-th switch position. This provides a temperature protection margin for that location. The junction temperature protection limit is set; the inspection range includes the calculation time in the pre-pulse segment and candidate sequence. The temperature margin is determined based on the one-sided underestimation error, reference temperature measurement error, and assembly differences verified offline within the applicable operating range, and the average fitting error cannot replace the underestimation risk. When exceeding the calibrated operating range, the original margin is not unconditionally extrapolated.

[0075] like Figure 2 As shown, when the second trajectory meets the protection constraints and the exit permit is valid, the first control step of the second trajectory is executed; when the second trajectory does not meet the constraints but the first trajectory does, the first control step of the first trajectory is executed, and both trajectories are regenerated and evaluated after new sampling arrives. If the second trajectory passes the prediction but the permit has not yet been obtained or has expired, the original exit action is not executed, but a re-evaluation is performed; the expired results of the first trajectory cannot be automatically reused.

[0076] When neither of the two tracks meets the protection constraints, or when the cumulative execution time of the first track since the current junction temperature warning is triggered reaches the preset waiting time limit, the normal exit process will be stopped and a fault handling request will be output. This stop does not mean that overmodulation will continue indefinitely; the fault handling will be taken over by the original protection path of the equipment, and verified actions will be performed according to the bus energy absorption capacity, speed and back EMF conditions. It is not stipulated that all gates must be turned off to achieve a safe state in the case of high-speed permanent magnet motors.

[0077] Exit Permit and Actual Loading Verification

[0078] like Figure 6As shown, after the second trajectory passes constraint verification, the permission management function 450 establishes a permission record, recording the verified first cycle parameters, target update event, and validity flag. The first cycle parameters include at least the counting mode, cycle value, phase comparison values, dead zone configuration, and output rules related to the candidate waveform; the target update event includes both the event type and cycle number. The permission is only for this first cycle and does not authorize direct execution in subsequent cycles.

[0079] Before loading, the controller 400 checks the loading parameters against the license record item by item and verifies that no target update events have been missed. Parameters can be written to the shadow register first, then the parameter version and event sequence number can be checked in the critical section, and finally, synchronous loading can be enabled. During this process, other ordinary control tasks are prohibited from overwriting related parameters without verification. Hardware fault outputs can still intervene with high priority and invalidate the original license.

[0080] If the parameters match and take effect as specified, load the initial cycle parameters and then mark the license as used. If the parameters are inconsistent, the target event has been missed, or the setup time is not expected to be met, do not enable the exit loading, revoke the license, and re-evaluate based on the latest active parameters. For edges that have already been loaded and output, do not claim that they can be reclaimed by revoking the license.

[0081] After loading, the actual activity parameters or corresponding loading completion records are read to update the actual state estimate, and the newly acquired phase current is compared with the predicted current at the same time. Parameter records are used to verify the execution at the timer level and are not equivalent to directly measuring the gate or junction temperature of each chip; drive faults are handled through the original drive diagnostics. If parameter inconsistencies or current deviations exceed limits, subsequent overmodulation exit steps are prohibited, and the evaluation is re-evaluated based on the actual loading records and new sampling.

[0082] In an illustrative operation scenario illustrating the control sequence, motor 500 is operating in an overmodulated range with limited bus voltage. The junction temperature at a certain switch position reaches the warning threshold but has not yet reached the emergency protection threshold. After sampling, controller 400 reads the timer state and confirms that the candidate parameters will only take effect in the subsequent synchronization update event. Therefore, it first predicts the changes in current and thermal state between sampling and that event.

[0083] Assuming that in this evaluation, the predicted junction temperature margin at a certain location exceeds the limit due to an increase in commutation events after boundary processing of the second trajectory, while the first trajectory still meets the protection constraints, then the first trajectory will be derated and resampled. This judgment is only for demonstration purposes and does not represent that the corresponding experimental results have been obtained. If the second trajectory meets the constraints in a subsequent evaluation, a permit will be generated that corresponds only to the first cycle and update event of that evaluation, and executed after pre-loading verification.

[0084] If the comparison parameters to be loaded have been changed during the verification, the exit permission will be invalidated, and the controller 400 will re-evaluate using the latest status instead of directly applying the previous temperature judgment to the changed waveform. If the parameters have been successfully loaded, but the deviation between the actual current and the prediction exceeds the limit, only subsequent exit steps will be prohibited and the status will be updated; the output pulses will still be included in the loss and thermal status as part of the existing operation process.

[0085] Parameter calibration and implementation verification, including device loss tables, thermal networks, and error margins, can be generated as follows: Initial parameters are established based on device data and actual driving conditions. Under actual heat dissipation installation conditions, heating and cooling processes are applied to the switch positions or bridge arms to calibrate the thermal response. If necessary, during the offline calibration phase, temperature-sensitive electrical parameters with established temperature relationships are used as junction temperature references, and the delay from shutdown to temperature measurement and its correction method are recorded. This reference temperature measurement chain is not a necessary sensor addition for online operation of this system.

[0086] The junction temperature underestimation was verified using loss step and cyclic loads not involved in parameter fitting, while simultaneously recording module temperature, actual load events, and phase current. Parameter applicability should be reassessed after changing module model, gate drive conditions, heatsink mounting method, or temperature sensing location. Routine maintenance focuses on temperature and current sampling calibration, heatsink installation consistency, and parameter version management.

[0087] During implementation, the worst-case execution time on the established controller should be determined, and the sum of the sampling transmission, dual-track prediction, register write, and synchronous load setup times should be verified to fall within the available time slots. When computational resources are insufficient, the prediction length can be reduced or a low-order model that has been error-verified can be used, but the pre-pulse segment and the first candidate cycle should still be covered; if the timing still cannot be met, the corresponding exit permission should not be granted.

[0088] The verification can be configured with a baseline based on real-time conditions, a processing method considering only load delay, a processing method with full boundary handling and permission checks, and a control group with boundary handling or permission binding removed respectively. Observe the consistency between the actual load and the evaluation object, boundary voltage changes, current prediction deviation, junction temperature underestimation, and calculation time.

[0089] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. An overmodulation protection system for motor inverters based on SiC device junction temperature estimation, characterized in that, The system includes a three-phase inverter bridge with SiC switching devices, a gate drive circuit, a sampling circuit, and a controller, wherein the controller controls the SiC switching devices through the gate drive circuit. The sampling circuit is connected to the controller to collect phase current, bus voltage, rotor position and module temperature; The controller updates the current junction temperature state of each SiC switch based on the sampling results and the loaded pulse sequence. When overmodulation is in operation and the junction temperature of any SiC switch reaches the warning threshold, a first trajectory that maintains overmodulation permission and reduces torque and a second trajectory that reduces the modulation limit and reduces torque are established. The pulse segment that cannot be changed by this update before the target loading time is taken as the pre-pulse segment shared by the two trajectories. The current state and junction temperature state at the end of the pre-pulse segment are predicted. The candidate pulse sequence connected with the pre-pulse segment is generated with the prediction result as the initial state. After performing dead-time and minimum pulse width processing based on the end switching state of the preceding pulse segment, the applied voltage and device loss are calculated using the same candidate pulse sequence to obtain the predicted current and predicted junction temperature. The corresponding trajectory is allowed to be executed only when the predicted current and predicted junction temperature meet the protection constraints. The exit permission of the second trajectory is bound to the first cycle pulse parameter of the candidate pulse sequence that has passed the verification and the target loading time. The exit permission is revoked when the loading parameter changes or the target loading time is missed.

2. The motor inverter overmodulation protection system based on SiC device junction temperature estimation according to claim 1, characterized in that, The controller includes a pulse width modulation timer; the target loading time corresponds to an update event in which the pulse width modulation timer makes the three-phase pulse parameters take effect synchronously; the preceding pulse segment is a pulse segment from the sampling reference time to the target loading time that has been determined by the loaded pulse sequence and cannot be changed by this modulation update.

3. The motor inverter overmodulation protection system based on SiC device junction temperature estimation according to claim 2, characterized in that, The controller recursively derives the current state and junction temperature state at the end of the pre-pulse segment based on the conduction interval and commutation event in the pre-pulse segment, combined with the phase current and the bus voltage, and records the conduction state, the most recent commutation time, and the remaining dead time of each SiC switching device at the end of the pre-pulse segment as common boundary conditions for generating the candidate pulse sequences corresponding to the two trajectories.

4. The motor inverter overmodulation protection system based on SiC device junction temperature estimation according to claim 3, characterized in that, The controller checks the boundary commutation of the candidate pulse sequence according to the common boundary conditions, and adjusts only the candidate pulses from the target loading time so that the commutation between the complementary SiC switching devices meets the preset dead time and the pulse duration meets the preset minimum pulse width. After adjustment, the applied voltage and the device loss are recalculated; for candidate pulse sequences that require changes to the pre-pulse segment to meet the preset dead time or the preset minimum pulse width, their corresponding trajectories are prohibited from execution.

5. The motor inverter overmodulation protection system based on SiC device junction temperature estimation according to claim 1, characterized in that, The controller stores a device loss parameter table and transient thermal network parameters; the device loss parameter table includes conduction loss parameters and commutation energy parameters corresponding to current, bus voltage, temperature and gate drive conditions; the controller calculates the device loss of the executed period based on the loaded pulse sequence, updates the current junction temperature state in combination with the module temperature and the transient thermal network parameters, and then recursively obtains the corresponding predicted junction temperature based on the device loss of each candidate pulse sequence.

6. The motor inverter overmodulation protection system based on SiC device junction temperature estimation according to claim 5, characterized in that, The controller includes a current regulator and a field weakening regulator. The controller uses the phase current, the rotor position, the integral state of the current regulator, and the limiting state of the field weakening regulator as inputs for current prediction. It performs current recursion based on the applied voltage corresponding to the candidate pulse sequence and performs the same current limiting, field weakening regulation, and anti-integral saturation processing as the actual control during the recursion process.

7. The motor inverter overmodulation protection system based on SiC device junction temperature estimation according to claim 6, characterized in that, The protection constraints include phase current limits, field weakening current limits, current tracking deviation limits, and device junction temperature limits. During the prediction period covering the preceding pulse segment and the subsequent candidate pulse sequence, the controller checks the current and junction temperature at each prediction moment and determines the temperature protection margin based on the junction temperature prediction error obtained from offline calibration within the corresponding operating range. Only when the phase current amplitude, field weakening current component, and tracking deviation relative to the corresponding current command of the predicted current satisfy the phase current limits, the field weakening current limits, and the current tracking deviation limits, respectively, and the predicted junction temperature of each SiC switching device, after being superimposed with the temperature protection margin, does not exceed the device junction temperature limit, is the corresponding trajectory deemed to satisfy the protection constraints.

8. The motor inverter overmodulation protection system based on SiC device junction temperature estimation according to claim 2, characterized in that, The controller establishes an authorization record for the first cycle pulse parameters and the target loading time corresponding to the second trajectory; before triggering the pulse loading corresponding to the update event, it checks the parameters to be loaded against the authorization record, and allows loading only when the checks are consistent and the target loading time is not missed, and invalidates the authorization record after loading; if the checks are inconsistent or the target loading time is missed, it does not perform an exit action and re-evaluates.

9. The motor inverter overmodulation protection system based on SiC device junction temperature estimation according to claim 7, characterized in that, When the second trajectory satisfies the protection constraint and the exit permission is valid, the controller executes the first control step of the second trajectory; when the second trajectory does not satisfy the protection constraint but the first trajectory does, the controller executes the first control step of the first trajectory, and regenerates and evaluates the two trajectories based on the sampling results after execution; when neither of the two trajectories satisfies the protection constraint, or when the cumulative execution time of the first trajectory since the current junction temperature warning is triggered reaches a preset waiting time limit, the controller prohibits the continued execution of the overmodulation exit process and outputs a fault handling request.

10. The motor inverter overmodulation protection system based on SiC device junction temperature estimation according to claim 9, characterized in that, After executing the first control step of the first trajectory or the second trajectory, the controller updates the current junction temperature state according to the actual loaded pulse parameters and compares the re-acquired phase current with the predicted current at the same time. When the actual loaded pulse parameters are inconsistent with the verified pulse parameters, or when the current comparison deviation exceeds the preset deviation limit, the subsequent overmodulation exit step is prohibited. The two trajectories are re-evaluated based on the actual loaded pulse parameters and the re-acquired phase current. When the junction temperature reaches the preset emergency protection threshold or when the two trajectories do not meet the protection constraints after re-evaluation, the fault handling request is output.