Wireless motor variable frequency speed regulation control method based on adaptive resonant frequency tracking and gain scheduling

Abstract

The application discloses a wireless driving motor variable frequency speed regulation control method based on adaptive resonant frequency tracking and gain scheduling, which reconstructs the frequency response of the primary impedance in real time by injecting a perturbation signal and locks the optimal working frequency; a variable frequency speed regulation module based on gain scheduling is used to establish the nonlinear mapping of the frequency offset and the secondary armature voltage to realize open-loop accurate speed regulation; a dynamic dead zone adjustment and global soft switching implementation mechanism is used to adaptively calculate the dead zone time according to the load current to ensure zero-voltage switching in all working conditions; and a multi-modal safety protection strategy based on frequency active retreat is used to realize physical layer energy suppression and fault isolation by using the gain attenuation characteristics of the resonant network in the detuning state. The application converts the complex closed-loop speed regulation into the physical mapping of frequency-voltage, and significantly improves the operation efficiency, dynamic response speed and intrinsic safety level of the system under the conditions of wide load range, variable coupling and severe electromagnetic environment.

Description

Technical Field

[0001] This invention relates to the field of wireless power transmission and motor drive control, specifically to a wireless drive motor variable frequency speed control method based on adaptive resonant frequency tracking and gain scheduling. Background Technology

[0002] Wireless drive motor systems achieve contactless energy transfer through high-frequency electromagnetic coupling, fundamentally eliminating the inherent problems of traditional mechanical contact power supply methods such as slip rings and cable chains, such as brush wear, electrical spark risk, and sealing difficulties. They show broad application prospects in fields with extremely high reliability requirements, such as aerospace, underwater equipment, medical devices, and semiconductor manufacturing.

[0003] Magnetic-coupled resonant wireless power transfer, as the current mainstream wireless power supply solution, is extremely sensitive to operating frequency in terms of power transfer efficiency—it can only maintain high-efficiency transmission within a very narrow bandwidth near the resonant point. However, existing wireless drive motor systems face the following four key technical bottlenecks in practical engineering applications:

[0004] (1) Resonant frequency drift leads to a sharp drop in efficiency across all operating conditions. The resonant frequency of a magnetically coupled resonant system... Self-inductance of the primary coil series compensation capacitor with primary side This is a joint decision. In actual operation, the switching of motor load from no-load to full-load will cause the equivalent impedance of the secondary side to change by several times or even tens of times. This is achieved through mutual inductance. Reflection impedance coupled to the primary side It will change accordingly. The change in the primary-side equivalent input impedance leads to The frequency response characteristics shift. Furthermore, thermal expansion of the coil due to temperature rise causes drift in self-inductance and mutual inductance parameters, and minute air gap changes caused by environmental vibrations also alter the coupling coefficient. The combined effect of these three factors causes the system's optimal resonant frequency to continuously deviate from the nominal value during actual operation, resulting in a sharp drop in efficiency for traditional fixed-frequency operation schemes when the load changes over a wide range. Experimental data shows that when the load switches from 10% to 100% of the rated load, the transmission efficiency of the fixed-frequency scheme can drop from the rated 91% to below 58%, an efficiency degradation of more than 33 percentage points.

[0005] (2) Communication link dependence constitutes a bottleneck for dynamic response. To achieve precise closed-loop speed regulation of the motor, traditional wireless drive systems install encoders or Hall sensors on the secondary side to obtain rotor position and speed information, and transmit the feedback data back to the primary controller via wireless radio frequency communication links such as Bluetooth, WiFi, or ZigBee. This dual-channel architecture of "power transmission + communication feedback" has two inherent defects: First, the communication delay is usually on the order of 10 to 100 milliseconds. When the motor load changes abruptly, the control loop cannot respond in time due to the delay, resulting in significant overshoot and oscillation of the speed, and in severe cases, even system instability. Second, in metal sealed cavities (Faraday cage effect), underwater high-attenuation media, or strong electromagnetic interference environments, the radio frequency signal is severely attenuated or even completely interrupted, and the communication reliability cannot be guaranteed. This makes the traditional solution face the risk of control failure in typical application scenarios such as deep-water thrusters and downhole drilling motors.

[0006] (3) Severe hard switching losses severely restrict the high-frequency and miniaturization of the system. Increasing the inverter operating frequency is the most direct technical approach to reduce the size of the magnetic coupling mechanism and increase the system power density. However, the switching losses of power devices... With operating frequency The relationship is approximately linear. Traditional fixed dead-time strategies cannot guarantee zero-voltage switching conditions across the entire load range—if the dead time is set too short, there is a risk of bridge arm shoot-through; if it is set too long, the parasitic capacitance cannot be fully charged and discharged under light load conditions, causing the switching transistor to be hard-turned on under non-zero voltage, resulting in severe switching losses and electromagnetic interference. This contradiction severely limits the further development of wireless drive systems towards megahertz-level high frequencies.

[0007] (4) Fault protection response suffers from a communication-dependent blind spot. Existing protection mechanisms mostly rely on secondary-side sensors to detect abnormal signals such as overcurrent and overtemperature, and then transmit these signals back to the primary-side controller via a communication link to execute a shutdown action. However, when the communication link is damaged simultaneously due to faults such as metal foreign object intrusion or coil short circuit, the protection command cannot be delivered to the primary-side power stage in a timely manner. The delay from the occurrence of the fault to the controller response can be as long as hundreds of milliseconds. During this window period, the power devices may have suffered irreversible damage, or even caused a safety accident. This architectural defect of "protection mechanism depending on the protected object" constitutes a major safety hazard in existing wireless drive systems.

[0008] Therefore, there is an urgent need for a wireless drive motor control method that can maintain high efficiency over a wide load range, achieve fast and accurate speed regulation without relying on communication links, enable soft-switching operation under all operating conditions, and possess physical-level intrinsic safety protection capabilities. Summary of the Invention

[0009] The purpose of this invention is to provide a wireless drive motor variable frequency speed control method based on adaptive resonant frequency tracking and gain scheduling. It aims to replace the reliance on wireless communication and high-computing-power controllers in the traditional architecture through a collaborative mechanism of adaptive frequency tracking, gain scheduling mapping, dynamic dead-time adjustment and active frequency pullback.

[0010] The present invention adopts the following technical solution:

[0011] A variable frequency speed control method for wireless drive motors based on adaptive resonant frequency tracking and gain scheduling is proposed. In a magnetically coupled resonant wireless power drive motor system, a wideband sweep frequency perturbation is superimposed on the power carrier to extract the amplitude and phase information of the primary side voltage and current, reconstruct the impedance frequency response curve, and update the current operating frequency through minimum value search and filtering. Then, the DC bus voltage after rectification and filtering of the secondary side is obtained based on the normalized frequency offset. The target armature DC bus voltage required is calculated according to the target speed command. The target operating frequency of the primary side inverter is calculated by reverse lookup using the gain scheduling function.

[0012] Furthermore, the specific steps to obtain the current operating frequency include:

[0013] (S210) Injection of broadband perturbation signal

[0014] A set of linear frequency modulated (FM) sweep perturbation signals is superimposed on a normal power transmission sinusoidal carrier wave, with a sweep range of [missing information]. The amplitude of the perturbation signal Power carrier amplitude Within 5%;

[0015] (S220) Carrier center alignment synchronization amplitude and phase acquisition

[0016] At each perturbation frequency point At this point, the dual-channel ADC is triggered to synchronously acquire the instantaneous voltage value across the primary resonant circuit. and instantaneous current value The sampling window length corresponds to at least one complete perturbation signal period; for the acquired time-domain data, the Goertzel algorithm is used to perform single-frequency discrete Fourier transform calculations at the frequency points. Calculate the voltage amplitude at the location Current amplitude and the phase difference between voltage and current ;

[0017] (S230) Reconstruction of the frequency response of the primary-side equivalent input impedance

[0018] Based on collection The amplitude and phase data at each discrete frequency sweep point are used to reconstruct the primary-side equivalent input impedance in real time according to the impedance definition. Discrete frequency response curves within the swept frequency range:

[0019]

[0020] (S240) Resonant pole localization and precise sub-frequency resolution localization

[0021] On the discrete impedance frequency curve reconstructed in step (S230), search for the impedance magnitude. The minimum point is found, and the sub-frequency resolution is accurately located between the found discrete minimum point and its two adjacent frequency points on the left and right.

[0022] (S250) First-order low-pass filter adaptive frequency update

[0023] The original optimal resonant frequency identified in step (S240) After smoothing, update to the system's current operating frequency:

[0024]

[0025] in, This is the operating frequency setpoint output to the PWM module for the current control cycle. The original optimal resonant frequency identified in the current frequency sweep cycle. This is the operating frequency of the previous control cycle. These are the filter coefficients.

[0026] Furthermore, the specific steps for obtaining the target operating frequency of the primary-side inverter include:

[0027] (S310) Calculation of normalized frequency offset

[0028] The normalized frequency offset is calculated using the following formula:

[0029]

[0030] in, The nominal resonant frequency specified at the factory. The optimal resonant frequency is tracked in real time during the current control cycle via step (S250);

[0031] (S320) Offline calibration and online table lookup of gain scheduling function

[0032] Based on the gain scheduling function covering the entire operating range, an online table is looked up using the normalized frequency offset. The DC bus voltage after secondary-side rectification and filtering is obtained through calculation or function approximation. ;

[0033] (S330) Target armature voltage feedforward calculation

[0034] According to the set target speed command The required target armature DC bus voltage is calculated using the back electromotive force equation of the permanent magnet synchronous motor:

[0035]

[0036] in, Let be the back electromotive force constant of the motor. The target mechanical angular velocity. To the current effective value of the load current Related comprehensive pressure drop compensation items;

[0037] (S340) Inverse lookup table of gain scheduling function and target operating frequency output

[0038] The target armature DC bus voltage calculated according to step (S330) In the gain scheduling function, a reverse lookup table operation is performed to find the target normalized frequency offset. ; Calculate the target operating frequency of the primary-side inverter based on the search results. .

[0039] Furthermore, it also includes system initialization and secure soft boot, the specific process of which is as follows:

[0040] (S110) End-to-end hardware authentication and communication self-test

[0041] The controller sends a query command to the driver chip of the digitally adjustable capacitor at the transmitting and receiving ends. If the controller receives a matching ID response and there is no fault status in the status register within the preset time window, it is determined that the full-link communication is normal and allows the subsequent startup process to proceed. If the self-test fails, the controller will forcibly pull down all 6 PWM outputs within a nanosecond time.

[0042] (S120) Neutral impedance safety parameter preset

[0043] The controller reads the nominal resonant frequency written during the factory calibration phase. The period register of the PWM timer of the primary inverter is initialized to the corresponding value, and the digital adjustable capacitor array at the transmitting end is also initialized. and receiver digitally adjustable capacitor array All are set to their middle code values;

[0044] (S130) Current closed-loop linear ramp soft start

[0045] Start the PWM module, set the initial duty cycle to a preset safe low threshold of 5%, and execute the current closed-loop ramp-up algorithm.

[0046] Furthermore, it also includes dynamic dead-zone adjustment and global soft switching, the specific process of which is as follows:

[0047] (S410) Real-time detection of the root mean square value of the load current

[0048] A current sampling resistor is connected in series between the source of the lower arm MOSFET and power ground of the primary side inverter bridge arm. The differential voltage signal across the resistor is amplified by an instrumentation amplifier with a fixed gain, and then sent to the controller through an anti-aliasing low-pass filter; continuous sampling is performed at a sampling rate synchronized with the control cycle; the controller samples the instantaneous current sequence in each control cycle. Perform root mean square calculation of the current effective value of the load current. , This represents the number of sampling points within a single period.

[0049] (S420) Dead-time adaptive feedforward calculation

[0050] Based on the current effective value of the load current Based on the switching characteristic parameters, the optimal dead time required for the current operating condition is calculated in real time according to the feedforward formula: ; The absolute minimum dead time to ensure safe straight-through of the bridge arm; This refers to the dead zone-current sensitivity coefficient.

[0051] (S430) Dead time and working frequency are coordinated for atomic updates

[0052] The dead time calculated in step (S420) Write the high-resolution dead-time configuration register of the PWM timer module to set the target operating frequency determined in step (S340). After being converted into a PWM timer period value, it is written into the corresponding register;

[0053] (S440) Real-time monitoring and closed-loop compensation of zero-voltage switching status

[0054] Real-time monitoring of the voltage at the midpoint of the primary-side inverter bridge arm and the gate drive signal of the switching transistor. The voltage level just before the rising edge arrives; if the voltage at the midpoint of the bridge arm... If the system has naturally transitioned to near-zero level through LC resonance, it is determined that the current ZVS condition is met and the system is operating normally. The controller collects the ZVS status flag in each control cycle. If the ZVS status is lost for three consecutive control cycles, the dead time compensation fine-tuning mechanism is automatically triggered.

[0055] Furthermore, it also includes multimodal security protection based on frequency active backoff, specifically including:

[0056] (S510) Real-time online calculation of efficiency residuals

[0057] Define the efficiency residual of the current control cycle as:

[0058]

[0059] in, The actual input power for the current control cycle, and the theoretical expected load power of the secondary side under the current operating conditions. ; This is the baseline value for expected efficiency.

[0060] (S520) Three-level threshold classification judgment for abnormal operating conditions

[0061] Level 1, Warning Threshold :when First time below At this time, the system records abnormal event logs and increases the efficiency residual monitoring frequency from once per control cycle to multiple sliding window average calculations within each control cycle, but the system maintains normal operation and does not trigger any power limiting measures.

[0062] Level 2, power reduction threshold :when consistently below Furthermore, if this state is maintained for more than 3 complete control cycles, the system determines that there is a non-transient, continuous abnormal energy loss, automatically triggers the power reduction protection mode, and enters the frequency active withdrawal process in sub-step S530.

[0063] Level 3, Emergency Stop Threshold :when Fall to When the bus current hardware overcurrent comparator is triggered, the system immediately performs a hardware-level emergency shutdown, directly blocking all 6 PWM output signals within nanoseconds.

[0064] (S530) Frequency active pullback physical layer energy suppression

[0065] When sub-step S520 determines that the power reduction protection mode has been triggered, the controller executes the frequency active pullback strategy, reducing the inverter operating frequency from the current resonant frequency. Quickly shift to a predefined safe frequency .

[0066] The beneficial effects of this invention are as follows:

[0067] (1) Frequency adaptive tracking enables efficient energy transfer under all operating conditions. This invention fundamentally solves the problem of efficiency drop caused by resonance offset when the load changes in the traditional fixed-frequency scheme by superimposing a broadband perturbation signal with an amplitude not exceeding 5% on the power carrier and using the Goertzel algorithm to reconstruct the impedance frequency response curve in real time. Experimental verification on a 500W / 85kHz prototype shows that when the load jumps from 10% to 100% of the rated load, the system efficiency of this invention remains above 87%, while the efficiency of the traditional fixed-frequency scheme drops sharply to below 58% under light load; when the air gap changes from 30 mm to 70 mm (simulating the relative displacement of the coil), the efficiency fluctuation range of this invention is only ±3%, while the efficiency fluctuation of the traditional scheme is as high as ±18%. The two comparisons fully verify the strong robustness of the frequency adaptive tracking mechanism to the dual disturbances of load change and position offset.

[0068] (2) Gain scheduling mapping achieves precise speed regulation without communication links. This invention utilizes an offline calibrated gain scheduling function. A deterministic nonlinear mapping between frequency offset and armature voltage is established, simplifying the closed-loop speed control link of "sensor → communication → controller" in the traditional scheme into a single-variable frequency regulation on the primary side. Experimental data shows that under the condition of a step increase in target speed from 500 rpm to 2500 rpm, the speed control response time of this invention (defined as the time required for the speed to reach 90% of the target value) is 82 milliseconds, and the steady-state speed error is less than 1.8%. In contrast, the traditional scheme using ZigBee communication (with a delay of about 20 milliseconds) for closed-loop PID speed control has a speed control response time of 185 milliseconds (the delay contributes about 65%), and the speed oscillates significantly when the communication is affected by electromagnetic interference. This method not only improves the speed control response speed by about 2.3 times, but is also completely unaffected by the quality of the communication link and the electromagnetic shielding environment.

[0069] (3) Dynamic dead-time adjustment ensures zero-voltage switching across the entire load range. This invention is based on the load current adaptive dead-time formula. With the frequency-dead-time coordinated atomic update mechanism, stable zero-voltage switching operation was achieved across the entire load range from 10% to 120% of the rated load. Thermal imaging measurements show that in systems using a fixed dead-time scheme, the inverter MOSFET junction temperature rises to 42 degrees Celsius (exceeding ambient temperature) under light load (20% load) conditions due to hard switching losses; while with the dynamic dead-time scheme of this invention, the junction temperature rise under the same conditions is only 18 degrees Celsius, a reduction of 57%. The average efficiency of the inverter across the entire load range increases from 91.2% with the fixed dead-time scheme to 95.8% with this invention, an efficiency improvement of 4.6 percentage points.

[0070] (4) Active frequency pullback achieves intrinsic safety protection at the physical level. This invention utilizes the voltage gain of the resonant network in the detuned state. The inherent physical characteristic of rapid attenuation with frequency shift allows for the physical suppression of energy transfer through active frequency shift when an efficiency residual abnormality is detected. Experimental verification shows that in a simulated fault scenario where a 20mm × 20mm × 2mm aluminum foreign object is inserted into the coil gap, the efficiency residual deteriorates from -3% to -18% within 1.2 milliseconds. After triggering the power reduction protection, the system shifts its operating frequency from 86.3 kHz to 92 kHz within 500 microseconds, the secondary output voltage drops sharply from 48 volts to 7.8 volts, the output power is suppressed to below 5% of the rated value, and the surface temperature of the foreign object reaches thermal equilibrium within 3 seconds. The physical delay of the entire protection response is only the frequency switching time (less than 500 microseconds), which is far superior to the sluggish protection response caused by communication delay (greater than 20 milliseconds) in traditional communication-based feedback schemes. Furthermore, it remains effective even in the extreme case of complete communication interruption, ensuring the inherent safety characteristics of the system from a physical mechanism perspective. Attached Figure Description

[0071] Figure 1 This is a block diagram of the overall architecture of the wireless drive motor variable frequency speed control system of the present invention.

[0072] Figure 2 The flowchart shows the algorithm for broadband impedance spectrum scanning and adaptive resonant frequency tracking in step S200.

[0073] Figure 3 This refers to the family of characteristic curves for gain-distributed variable frequency speed regulation in step S300.

[0074] Figure 3 In the figure, (a) is the normalized voltage gain curve of the resonant network under different normalized frequency offsets; (b) is the mapping curve of the gain scheduling function; and (c) is the linear relationship curve between the target motor speed and the target armature voltage.

[0075] Figure 4 This is a schematic diagram of the key waveforms for dynamic dead-zone adjustment and zero-voltage soft switching in step S400.

[0076] Figure 5 The figure shows the experimental curves comparing the overall system efficiency of the traditional fixed-frequency scheme and the frequency adaptive scheme of this invention.

[0077] Figure 6 The waveform diagram shows the dynamic response comparison between the traditional ZigBee communication closed-loop PID speed control method and the gain scheduling open-loop speed control method of this invention when the motor load changes abruptly.

[0078] Figure 6 In the figure, (a) is a comparison chart of rotational speed response, (b) is a comparison chart of speed error, and (c) is a chart of real-time changes in the normalized frequency offset of the present invention.

[0079] Figure 7 The dynamic response curve of the speed for open-loop speed regulation with gain control.

[0080] Figure 8 Comparison of timing waveforms for adaptive dead-time ZVS switching.

[0081] Figure 9 The waveform diagram shows the dynamic test results of multimodal safety protection under foreign object intrusion fault. Detailed Implementation

[0082] The integrated adaptive control scheme of "tracking-scheduling-soft switching-protection" proposed in this invention aims to replace the dependence of traditional architectures on wireless communication and high-computing-power controllers through a collaborative mechanism of frequency adaptive tracking, gain scheduling mapping, dynamic dead-time adjustment and active frequency withdrawal.

[0083] The hardware foundation for the control and operation of this invention includes:

[0084] (1) Primary side: Includes the primary DC power supply and the primary inverter (full-bridge or half-bridge), used to generate a frequency-adjustable high-frequency AC power signal. A compensation capacitor is connected in series at the output of the primary inverter. Primary-side transmitting coil and its equivalent resistance This forms the primary resonant circuit.

[0085] (2) Secondary side: including the secondary receiving coil and compensation capacitor. The rectifier and filter unit and the motor load are connected in series to form the secondary energy receiving and driving circuit.

[0086] (3) Sampling and Control: Primary voltage and current sampling unit, used to collect electrical parameters of the primary resonant circuit or bus side. Digital controller, used to perform frequency sweep perturbation, Goertzel amplitude and phase extraction, impedance curve reconstruction, gain scheduling lookup, dead time calculation and protection judgment. PWM drive and hardware protection unit, used to adjust the operating frequency, duty cycle and dead time of the primary inverter, and to block the PWM output in case of abnormality.

[0087] Preferably, the primary-side compensation capacitor and secondary side compensation capacitor A digitally adjustable capacitor array (including a driver chip) is selected for initializing or calibrating the bilateral compensation network. Alternatively, a bridge arm switching node voltage comparator is provided to detect zero-voltage switching states.

[0088] The magnetically coupled resonant wireless power supply main topology in the above hardware foundation is a common basic architecture in this field; the present invention is based on this basic architecture and configures and executes specific frequency tracking, gain scheduling, soft switching and safety protection control processes.

[0089] This invention can be applied to the previously proposed communication-free wireless drive motor system based on physical layer adaptation and impedance pass-through (Chinese patent application 2026110628804), and can also be applied to other magnetically coupled resonant wireless drive motor platforms.

[0090] Figure 1 This is a schematic diagram of the overall framework of the present invention. Figure 1 The main power transmission link in the system, namely "primary-side inverter - SS compensation magnetic coupling network (magnetic coupling resonant mechanism) - secondary-side rectification and filtering - motor load", is a common basic framework in magnetic coupling resonant wireless drive systems.

[0091] The digital controller employs a dual-core DSP. Primary-side voltage and current sampling channels are connected to the digital controller's ADC port for impedance spectrum reconstruction and efficiency residual calculation. The digital controller controls the inverter's operating frequency, duty cycle, and dead time via a PWM module. The digital controller directly blocks the PWM output via a hardware Trip Zone or equivalent protection port. The digital controller connects to the driver chip of the digitally adjustable capacitor array via a digital interface to preset or adjust compensation network parameters. A bridge arm midpoint voltage detection unit is connected to the digital controller for ZVS status identification and dead-time compensation.

[0092] Step S100 System initialization and secure soft boot.

[0093] The core objective of this step is to build an electrically safe and state-controlled system startup environment to prevent overcurrent surges, voltage stress damage, and inverter arm shoot-through risks caused by severe mismatch of resonant parameters at the moment of power-on.

[0094] (S110) Full-link hardware authentication and communication self-test.

[0095] The controller (DSP or FPGA) sends a query command (WHO_AM_I) to the driver chip of the digitally adjustable capacitor at both the transmitting and receiving ends via an isolated SPI bus. This command is a standard slave device identification register read operation. If the controller receives a matching ID response within a preset 100-microsecond time window and there are no fault states such as over-temperature (OT) or undervoltage lockout (UVLO) flags in the status register, the entire link communication is considered normal, and the system is allowed to proceed to the subsequent startup process. If the self-test fails (no response received within timeout, ID mismatch, or presence of fault flags), the controller forcibly pulls all six PWM outputs low within a nanosecond time using the hardware TRIP_ZONE pin, completely blocking the energy transmission channel at the physical level and preventing the power stage from being accidentally turned on under abnormal communication conditions.

[0096] (S120) Neutral impedance safety parameter preset.

[0097] The controller reads the nominal resonant frequency written during the factory calibration phase from the built-in non-volatile memory (EEPROM or FLASH). (In this embodiment) And initialize the inverter PWM timer's period register to the corresponding value. Simultaneously, set the transmitter's digitally adjustable capacitor array... and receiver digitally adjustable capacitor array All values ​​are set to their intermediate code values ​​(e.g., for an 8-bit resolution 256-level capacitor array, the code value is set to 128). "Neutral impedance state" specifically refers to the state where both sides of the resonant circuit exhibit weak inductive or purely resistive impedance characteristics through capacitor array adjustment. The physical significance of this preset strategy is that at startup, if the resonant network exhibits capacitive load characteristics, the voltage phase at the midpoint of the inverter bridge arm will lead the current, causing the switching transistors to experience extremely high voltage stress under hard switching conditions; while the weak inductive or purely resistive neutral impedance state ensures that the inverter faces a relatively safe switching environment under initial conditions.

[0098] (S130) Current closed-loop linear ramp soft start.

[0099] The PWM module is started, with the initial duty cycle set to a preset safety low threshold of 5%. A current closed-loop ramp-up algorithm is executed: the PWM duty cycle is increased at a fixed linear slope of 1% of the duty cycle every millisecond until the rated operating duty cycle is reached (45% in this embodiment, corresponding to the normal operating point of the full-bridge inverter). Throughout the ramp-up process, dual protection monitoring is implemented. The first layer is current rate of change protection: real-time calculation of the bus current growth slope. If the current exceeds the preset motor starting surge threshold (calibrated offline based on the motor's rated current and moment of inertia), the duty cycle increment is immediately paused and maintained for 10 milliseconds. It continues to increase only after the current has decreased and stabilized. This mechanism effectively suppresses the surge current impact at the moment of motor startup. The second layer is cycle-by-cycle overcurrent protection: if the instantaneous value of the bus current exceeds the hardware overcurrent protection threshold... Within the current PWM cycle, the hardware Trip Zone module blocks all PWM outputs within nanoseconds, achieving cycle-by-cycle hardware current limiting protection.

[0100] Step S200: Wideband impedance spectrum scanning and adaptive resonant frequency tracking.

[0101] Figure 2 shows the complete algorithm flowchart for step S200, wideband impedance spectrum scanning and adaptive resonant frequency tracking. The signal processing link is built based on prototype calibration values ​​such as a nominal resonant frequency of 85kHz, a 10ms sweep period, and a ±15% sweep interval. The process begins by superimposing a linear frequency-modulated perturbation signal with an amplitude not exceeding 5% of the carrier frequency onto a normal power carrier. Then, the PWM is configured in center-aligned mode, and the instantaneous values ​​of the primary-side voltage and current are synchronously acquired by a dual-channel ADC at the midpoint of the switching edge to avoid switching ringing noise. Next, the amplitude and phase data are extracted frequency-by-frequency using the Goertzel algorithm to reconstruct the discrete frequency response curve of the primary-side equivalent input impedance in real time. Then, the minimum modulus value is searched on the impedance curve as the resonant base point, and combined with quadratic parabolic interpolation to achieve high-precision sub-frequency point positioning of the optimal resonant frequency. Finally, a first-order low-pass digital filter with a time constant of 50ms is used to smooth the frequency values, and the updated optimal operating frequency is sent to the PWM period register, completing one round of resonant tracking and returning to the main control loop. Figure 2 The system uses solid lines to distinguish data / signal flow and dashed lines to distinguish control flow, fully restoring the timing, algorithm parameters, and hardware execution logic of the entire process of perturbation injection, sampling, impedance reconstruction, pole optimization, and filter update. All frequency sweep ranges, sampling timings, and filtering parameters match the actual operating values ​​of the system.

[0102] This step is the core foundation of the frequency adaptive mechanism of this invention. Its technical idea is that, rather than relying on a fixed-parameter model calibrated offline, it dynamically locks the optimal resonant frequency of the system under the current operating conditions by actively injecting a broadband perturbation signal and analyzing the frequency response characteristics of the primary impedance in real time. The following section elaborates on the five sub-steps of this step, starting from the mathematical and physical model.

[0103] (S210) Injection of broadband perturbation signal.

[0104] Sinusoidal carrier with normal power transmission Above, a set of linear frequency modulated (Chirp) sweep perturbation signals are generated and superimposed through the digital frequency synthesizer (DDS) inside the DSP. The sweep range is... That is, under the nominal resonant frequency of 85kHz in this embodiment, the sweep frequency range is 72.25kHz to 97.75kHz, and the sweep frequency period is... The amplitude of the perturbation signal Strictly control the power carrier amplitude Within 5%, that is This amplitude constraint ensures that the impact of perturbation injection on the energy efficiency of power transmission is negligible (the additional power contributed by the perturbation signal does not exceed 0.25% of the rated power, i.e. The perturbation signal injection strategy makes full use of the narrowband bandpass filtering physical characteristics of the resonant network itself: frequency components far from the current resonant operating point are naturally attenuated by the frequency selectivity of the resonant circuit, and only frequency components close to the resonant region can generate observable voltage and current responses in the circuit, thereby effectively extracting the frequency response characteristics of the system without significantly interfering with power transmission.

[0105] (S220) Carrier center alignment synchronization amplitude and phase acquisition.

[0106] To achieve high signal-to-noise ratio signal acquisition, this sub-step employs a hardware-level timing optimization strategy. The PWM timer is configured in center-aligned mode. In this mode, the PWM counter increments from zero to the period value and then decrements back to zero, forming a symmetrical triangular wave carrier. The ADC sampling trigger time is precisely set at the midpoint of the PWM counter's overflow / underflow, i.e., the moment when the counter value equals half the period value. This sampling timing is based on a clear physical principle: in center-aligned PWM mode, the inverter bridge arm switching occurs when the counter reaches zero and the period value, and the midpoint is precisely located in the middle of the two switching edges, at which point the switching ringing has sufficiently attenuated to a negligible level. At each perturbation frequency point... At this point, the dual-channel ADC is triggered to synchronously acquire the instantaneous voltage value across the primary resonant circuit. and instantaneous current value The sampling window length corresponds to at least one complete perturbation signal period. For the acquired time-domain data, the Goertzel algorithm is used to perform single-frequency discrete Fourier transform calculations. This algorithm has a significant computational efficiency advantage over the full FFT when processing a small number of specific frequency points. At the frequency points... Calculate the voltage amplitude at the location Current amplitude and the phase difference between voltage and current .

[0107] (S230) Reconstruction of the frequency response of the primary side equivalent input impedance.

[0108] Based on collection The amplitude and phase data at each discrete swept frequency point are used to reconstruct the discrete frequency response curve of the primary-side equivalent input impedance within the swept frequency range in real time, according to the impedance definition:

[0109]

[0110] This impedance curve has a clear physical meaning; it is a comprehensive mapping of the primary-side resonant network's own impedance and the secondary-side reflected impedance. For the SS double-sided series compensated topology, the analytical expression for the primary-side equivalent input impedance is:

[0111]

[0112] in, This is the equivalent series resistance of the primary coil. The self-inductance of the primary coil This is a primary-side series compensation capacitor. Angular frequency. Reflection impedance. The expression is:

[0113]

[0114] In the formula, The mutual inductance between the primary and secondary coils, The total impedance of the secondary circuit is:

[0115]

[0116] in, This is the equivalent series resistance of the secondary coil. To calculate the equivalent load resistance referred to the secondary AC side (for scenarios driving permanent magnet synchronous motors), (Dynamically changes with the operating conditions of the motor) For the self-inductance of the secondary coil, This is a secondary-side series compensation capacitor. Analysis of the above formula shows that when the load... When changing, Change, and then Change ultimately leads to The amplitude and phase frequency characteristics of the resonant frequency both shift, which is the physical reason why step S200 needs to track the resonant frequency shift in real time.

[0117] (S240) Resonant pole location and sub-frequency resolution precise location.

[0118] On the discrete impedance frequency curve reconstructed in sub-step S230, search for the impedance magnitude. The minimum point. According to the basic principle of resonant circuits, in the series resonance state, the reactance components of the circuit cancel each other out (i.e., ...). imaginary part and The sum of these values ​​is zero, and the input impedance exhibits purely resistive behavior and reaches a local minimum. Therefore, The frequency corresponding to the minimum point is the current coupling coefficient. and load Optimal resonant frequency under certain conditions To improve frequency positioning accuracy and overcome the limitation of the discrete interval of frequency sweep points, a quadratic parabolic interpolation method is used for precise sub-frequency resolution positioning between the searched discrete minimum points and their two adjacent frequency points. This involves fitting a quadratic parabola to three discrete points. The vertex of the parabola corresponds to a more precise resonant frequency value within the continuous domain.

[0119] (S250) First-order low-pass filter adaptive frequency update.

[0120] The original optimal resonant frequency identified in sub-step S240 After smoothing using a first-order low-pass digital filter (exponentially weighted moving average, EWMA), the system's current operating frequency is updated.

[0121]

[0122] in, This is the operating frequency setpoint output to the PWM module for the current control cycle. The original optimal resonant frequency identified in the current frequency sweep cycle. This is the operating frequency of the previous control cycle. These are the filter coefficients. With control cycle and filter time constant The relationship is .

[0123] In this embodiment, the control cycle Filter time constant Calculated Choose a larger time constant (much larger than the sweep period). The purpose of this filter is to effectively suppress random frequency jitter caused by noise interference during a single frequency sweep measurement, making the system's frequency update process smooth and stable, and preventing high-frequency oscillations near the resonant point. Simultaneously, the bandwidth of this filter is sufficient to track the resonant frequency drift caused by load changes—the time constant of motor load changes is typically on the order of tens of milliseconds, far exceeding the filter's response time of 50 milliseconds.

[0124] Step S300: Variable frequency speed control based on gain scheduling.

[0125] This step establishes a deterministic mapping relationship between the resonant frequency offset and the secondary DC bus voltage (and consequently, the motor speed), achieving open-loop precise speed regulation that completely eliminates the secondary communication feedback link. Its core idea lies in the fact that the magnetically coupled resonant network is essentially a bandpass energy transmission channel that is highly sensitive to frequency, and its voltage gain... The attenuation characteristic exhibits a regular and reproducible pattern as the operating frequency deviates from the resonant point. By establishing a quantitative mapping of this attenuation characteristic through offline calibration, the secondary output voltage can be precisely controlled online by adjusting the primary-side operating frequency.

[0126] (S310) Calculation of normalized frequency offset.

[0127] Let the nominal resonant frequency specified by the manufacturer be... The current optimal resonant frequency output after impedance spectrum scanning and first-order low-pass filtering in step S250 is: (i.e., in step S250) This value already includes compensation for resonant point offset caused by current load, coupling state, and temperature drift, defining the normalized frequency offset. .

[0128] The normalized frequency offset It is a dimensionless quantity that comprehensively reflects the combined effects of load changes, coil relative displacement, and ambient temperature drift on the system's resonant state. The advantage of using the normalized form is that systems with different power levels and nominal frequencies can use a unified gain scheduling function, and can be ported simply by recalibrating.

[0129] Figure 3 (a) With the normalized frequency offset as the horizontal axis and the normalized resonant voltage gain as the vertical axis, the curve reaches the peak gain of 1 at the offset point 0 and decreases monotonically with the frequency offset, covering the effective working range of -3% to +5%.

[0130] (S320) Offline calibration and online table lookup of gain scheduling function.

[0131] During the system factory calibration phase, a gain scheduling function was established through offline experiments covering the entire operating range. Specifically, this was conducted at load rates from 10% to 120% of rated load and coupling coefficients... Within the complete two-dimensional working domain, with a sufficiently fine mesh (e.g. A system scan experiment was conducted with increments of 0.1%, and the DC bus voltage after secondary rectification and filtering at each operating point was recorded. This forms a two-dimensional mapping table. In practical engineering implementation, this mapping relationship is simplified into a single-variable function. .function A hybrid approach combining piecewise linear interpolation lookup tables (LUTs) with local third-order polynomial fitting is employed. The effective range is evenly divided into multiple sub-intervals, each storing the linear slope and intercept parameters. During online table lookup, based on the current... Extract the corresponding linear coefficients from the interval index for fast calculation. For applications requiring higher precision, the coefficients of a third-order polynomial can be stored in each sub-interval to achieve smooth and high-precision function approximation. Within its effective operating range, it exhibits a monotonically decreasing mathematical property (the further away from the resonant point, the lower the energy transfer gain of the resonant network and the smaller the secondary output voltage). This monotonicity ensures that the subsequent reverse lookup table has a unique solution.

[0132] Figure 3 (b) is the gain scheduling mapping curve. The secondary DC voltage decreases synchronously as the frequency offset increases. Using 0.1% step discrete node segmented interpolation, the target frequency offset can be uniquely solved in reverse based on the target armature voltage.

[0133] (S330) Target armature voltage feedforward calculation.

[0134] According to the user-defined target speed command The required target armature DC bus voltage is calculated using the back electromotive force equation of the permanent magnet synchronous motor:

[0135]

[0136] in, The back electromotive force constant of the motor (unit: volts per radian per second or volts per revolution per minute, depending on...) (Unit selection) The target mechanical angular velocity. To the current effective value of the load current The relevant comprehensive voltage drop compensation term, its physical components include the forward voltage drop of the rectifier bridge diodes. (Typically 0.7 volts to 1.5 volts per diode, full-bridge rectifier includes two paths), equivalent series resistance of the circuit. Ohmic voltage drop and the equivalent voltage drop of copper losses in the motor armature winding. .because Real-time measurement can be performed via step S410. Online calculations can be performed using pre-calibrated lookup tables or linear fitting methods. This feedforward compensation mechanism effectively eliminates the steady-state error in speed regulation caused by voltage drops in lines and components, achieving high-precision open-loop speed control.

[0137] (S340) Gain scheduling function reverse lookup table and frequency output.

[0138] The target armature voltage calculated based on sub-step S330 In the calibrated gain scheduling function Perform a reverse lookup operation to find the table that makes Established target normalized frequency offset .because Within its effective working range, the expression is strictly monotonically decreasing, and reverse lookup has a mathematical guarantee of uniqueness. The lookup algorithm can employ binary search or direct reverse interpolation based on interval positioning.

[0139] Calculate the target operating frequency of the inverter based on the search results:

[0140]

[0141] The target frequency value As a new frequency setting value for the inverter's PWM module, it is directly output and takes effect within the current control cycle (100 microseconds). Compared with the traditional closed-loop speed control link of "ADC sampling → communication transmission → PID calculation → PWM update", the speed control response delay of this method is only the calculation time of a single control cycle, which is greatly reduced from tens of milliseconds in the traditional scheme to the order of hundreds of microseconds, achieving a response speed improvement of more than two orders of magnitude.

[0142] Figure 3 (c) The target motor speed and target armature voltage are approximately linearly related, with the slope matching the motor back EMF constant of 0.015V / rpm, and the load voltage drop compensation term caused by rectification and winding is superimposed. Figure 3 The operating logic and calibration parameters of open-loop speed regulation without secondary side communication, which fully restores the "speed command - target voltage - frequency offset - inverter operating frequency", are restored.

[0143] Step S400: Dynamic dead zone adjustment and global soft switching are implemented.

[0144] This step ensures that the inverter power devices operate stably in a zero-voltage switching state under full operating frequency range and full load conditions through load current adaptive dead-time calculation and frequency-dead-time collaborative update mechanism, thereby minimizing switching losses, reducing electromagnetic interference and improving system reliability.

[0145] (S410) Real-time detection of the root mean square value of the load current.

[0146] The primary-side current sampling unit includes a low-resistance sampling resistor connected in series between the source of the lower arm of the primary-side inverter bridge and power ground, a differential amplifier connected across the sampling resistor, an anti-aliasing low-pass filter, and the ADC input of the digital controller. The controller calculates the effective current value based on this sampled signal, which is used for adaptive dead-time calculation, soft-start current limiting, and safety protection determination.

[0147] A high-precision, low-temperature-coefficient current sampling resistor is connected in series between the source of the MOSFET in the lower arm of the primary inverter bridge and the power ground. (Typical values ​​range from 10 milliohms to 50 milliohms). The differential voltage signal across the resistor is amplified by an instrumentation amplifier (INA series or similar devices) with a fixed gain, and then fed into the DSP's ADC module through an anti-aliasing low-pass filter. The ADC continuously samples at a sampling rate of 20 kHz, synchronized with the control cycle. The DSP processes the acquired instantaneous current sequence within each control cycle. Perform root mean square calculation:

[0148]

[0149] in, This represents the number of sampling points within a single period.

[0150] (S420) Dead time adaptive feedforward calculation.

[0151] Based on the current effective value of the load current Based on the switching characteristic parameters of the power devices used, the optimal dead time required for the current operating condition is calculated in real time using the feedforward formula: .

[0152] In a zero-voltage switching (ZVS) topology, after the switching transistor is turned off, the midpoint voltage of the bridge arm affects the parasitic output capacitance of the upper and lower transistors through the inductor current of the resonant circuit. The time required for the switch to charge and discharge naturally to transition to zero level or the bus voltage level (depending on the switch position) is the optimal dead time required to achieve ZVS. When the load current When the load current is small, the parasitic capacitance charges and discharges slowly, requiring a long dead time. As the load current increases, the energy stored in the inductor increases, accelerating the charging and discharging process and correspondingly shortening the required dead time. The parameters in the formula... The absolute minimum dead time required to ensure safe shoot-through of the bridge arm depends on the turn-off delay time of the power devices. The difference in propagation delay between the drive circuit and the device is typically 50 to 80 nanoseconds for silicon carbide (SiC) MOSFET devices and 15 to 30 nanoseconds for gallium nitride (GaN) HEMT devices. (Scale factor) The dead zone-current sensitivity coefficient is obtained by experimentally measuring the optimal dead time of ZVS under different load current conditions and performing least squares linear regression fitting on the measurement data. Its physical dimension is time divided by current.

[0153] (S430) Dead time and working frequency are updated in tandem with atomic updates.

[0154] The dead time calculated in sub-step S420 Write the high-resolution dead-time configuration register of the PWM timer module (corresponding to the HRDEAD register in the TI C2000 series DSP, with a dead-time resolution of up to 150 picoseconds), and set the target operating frequency determined in step S300. After being converted into PWM timer period values, the values ​​are written to the corresponding register (TBPRD register). Utilizing the PWM module's global shadow register loading mechanism (which ensures that the value written to the shadow register is only synchronously copied to the active register at the specific moment the PWM counter returns to zero), the frequency and dead-time values ​​atomically take effect simultaneously at the start boundary of the same PWM cycle. This "atomic update" mechanism fundamentally eliminates the intermediate transition states that may be caused by asynchronous updates of frequency and dead-time, avoiding the risk of instantaneous bridge arm shoot-through or ZVS loss due to a change in frequency but not in dead-time.

[0155] (S440) Real-time monitoring and closed-loop compensation of zero-voltage switch status.

[0156] By using an auxiliary high-speed voltage comparator circuit, the voltage at the midpoint of the primary-side inverter bridge arm (i.e., the voltage of the switching node SW-node) is monitored in real time against the gate drive signal of the switching transistor. The voltage level just before the rising edge arrives. If the voltage at the midpoint of the bridge arm... It has naturally transitioned to near-zero level through LC resonance, and the specific judgment criteria are as follows: (i.e., below the bus voltage) If the dead time is within 5%, the current ZVS condition is considered met, and the system operates normally. The DSP collects the ZVS status flag in each control cycle. If the ZVS status is lost for three consecutive control cycles (i.e., within 300 microseconds), the dead time compensation fine-tuning mechanism is automatically triggered: the dead time is adjusted in a minimum step increment of 1 to 2 nanoseconds each time, and the ZVS recovery status in subsequent cycles is monitored to form an online closed-loop self-calibration of the ZVS status.

[0157] The bridge arm midpoint voltage detection unit includes a high-impedance voltage divider network, a limiting protection circuit, a high-speed voltage comparator, and a level conversion interface. The input of the high-impedance voltage divider network is connected to the bridge arm midpoint SW-node of the primary-side full-bridge inverter, and the output of the voltage divider is connected to the first input of the high-speed comparator. The second input of the comparator is connected to a reference threshold, which is set to the DC bus voltage V. dc The voltage is 5% of the reference threshold or the corresponding voltage divider value. When the SW-node voltage is lower than the reference threshold before the rising edge of the gate drive arrives, the comparator outputs a ZVS satisfy signal; otherwise, it outputs a ZVS loss signal. The comparator output is connected to the controller GPIO, capture input, or Trip Zone auxiliary input for dead-time closed-loop fine-tuning and protection triggering.

[0158] Figure 4To compare the timing waveforms of dynamic dead-time adjustment and global zero-voltage soft switching, the test conditions were divided into three categories: 20% light load, 100% rated load, and 120% heavy load. For each group, the gate drive signals of the traditional fixed dead-time scheme and the adaptive dead-time scheme of this invention were compared synchronously. Midpoint voltage of bridge arm The dead zone of this scheme follows The adaptive formula automatically extends to approximately 54ns under light load and adjusts to 91ns under full load. Using 5% of the bus voltage as the ZVS threshold, the measured voltage of the bridge arm before the switch is turned on under all operating conditions is 4% lower than the bus voltage (minimum 1.8V), stably achieving zero-voltage turn-on. In contrast, the fixed dead-time scheme has insufficient dead time under light load, resulting in hard turn-on voltage spikes. This clearly demonstrates the effect of the adaptive dead-time mechanism in achieving low-loss operation under full load based on the measured parameters of the prototype machine.

[0159] Step S500 Multimodal security protection based on frequency active backoff.

[0160] This step constructs a physical-level multimodal safety protection system that does not rely on secondary sensors or communication links. Its core technical idea lies in utilizing the inherent physical characteristic of a magnetically coupled resonant network where voltage gain rapidly decreases in a detuned state. Upon detecting abnormal operating conditions, it physically suppresses energy transmission capability by actively shifting the operating frequency, achieving intrinsic safety protection that "replaces software logic with physical laws." This strategy fundamentally eliminates the architectural paradox of "protection mechanisms depending on communication links" in traditional solutions.

[0161] (S510) Real-time online calculation of efficiency residuals.

[0162] Based on the primary DC bus voltage and bus current Calculate the actual input power for the current control cycle based on real-time measurements:

[0163]

[0164] Meanwhile, based on the mutual inductance parameters identified in real time in step S200 (Through reflection impedance) (obtained through model inversion) and current load estimates Calculate the theoretical expected load power of the secondary side under the current operating conditions. Furthermore, the efficiency residual of the current control cycle is defined as:

[0165]

[0166] in, This is the expected efficiency baseline value obtained by looking up a table based on historical efficiency data from long-term system operation and current operating conditions (including output power level, ambient temperature, operating time, etc.). Under normal operating conditions, It is usually in the range of 85% to 93%. When A significant deviation from zero (especially in the negative direction) indicates the presence of non-load-related abnormal energy loss in the system. Typical sources of this abnormality include: eddy current loss caused by metal foreign objects intruding into the coil gap, dielectric loss caused by the breakdown of the compensation capacitor, and local hot spot loss caused by short circuits between coil turns.

[0167] Figure 5 This is a comparison curve of the overall system efficiency of the traditional fixed-frequency scheme and the adaptive resonant tracking scheme of this invention under a standard 50mm air gap coil. The horizontal axis covers the rated load range of 10% to 120%, and the vertical axis represents the overall efficiency from the DC bus to the motor shaft. The traditional fixed-frequency scheme only achieves a peak efficiency of 91% under rated load, and the efficiency drops below 58% under 10% light load conditions, with an efficiency difference of over 33 percentage points across the entire range. This invention, relying on an impedance scanning resonant tracking algorithm with a period of 10ms, maintains a stable efficiency of over 87% across the entire load range, with efficiency fluctuations of only ±3%. The curve is smooth with no significant attenuation. All efficiency differences and load range values ​​are from prototype machine comparison experiments, intuitively demonstrating the operational advantages of adaptive frequency tracking in suppressing resonant drift and maintaining high efficiency over a wide load range.

[0168] (S520) Three-level threshold classification judgment for abnormal working conditions.

[0169] Taking into account both system safety margin and engineering practicality, a three-tiered progressive safety threshold is set:

[0170] Level 1, Warning Threshold :when First time below At that time, the system records an abnormal event log (including timestamp, current operating parameters, and...). This threshold (value) increases the efficiency residual monitoring frequency from once per control cycle to multiple sliding window averaging calculations within each control cycle, while maintaining normal system operation and not triggering any power limiting measures. This threshold is used to capture early, subtle efficiency anomalies.

[0171] Level 2, power reduction threshold :when consistently below If this state is maintained for more than 3 complete control cycles (i.e. 300 microseconds), the system determines that there is a non-transient continuous abnormal energy loss, automatically triggers the power reduction protection mode, and enters the frequency active pullback process in sub-step S530.

[0172] Level 3, Emergency Stop Threshold :when Fall to When the bus current hardware overcurrent comparator is triggered (without any software judgment), the system immediately performs a hardware-level emergency shutdown—the PWM Trip Zone module directly blocks all 6 PWM output signals within nanoseconds.

[0173] (S530) Frequency active pullback physical layer energy suppression.

[0174] When sub-step S520 determines that the power reduction protection mode has been triggered, the controller executes the frequency active rollback strategy, setting the inverter's current operating frequency... Switch to a predefined safe frequency ,

[0175]

[0176] in, The current optimal resonant frequency determined in step S250. This is the frequency offset by which the voltage gain of the resonant network drops to a safe range. Typically, under normal operating conditions... When retracting, the current optimal resonant frequency is used as a reference to shift towards the low gain region.

[0177] Safety frequency offset The selection principle is based on the voltage gain function of the resonant network. The frequency response characteristics. For the magnetically coupled resonant network with SS compensation topology, its normalized voltage gain... At the resonance point The maximum value of 1.0 is taken at this point, and then... The increase of exhibits a monotonically decreasing trend. A complete gain-frequency response curve was measured through offline experiments, and a selection was made that... frequency offset For quality factors A typical resonant circuit with a value of 50 to 100, Approximately The voltage is 5% to 8% of the rated torque. Operating at this safe frequency, the secondary-side rectified output voltage is physically suppressed to a critical level (typically 5% to 10% of the rated torque) that just keeps the secondary-side control circuit from losing power and allows the motor output torque to just overcome static friction. This state is called "hot standby safe state," and its advantage is that the system is not completely shut down. After external abnormal factors are eliminated (such as the removal of foreign objects), full-power operation can be quickly restored within a single control cycle. When sub-step S520 triggers a hardware emergency stop, the Trip Zone module bypasses all software judgment and communication links, directly blocking the PWM output within nanoseconds to ensure the ultimate speed of fault response.

[0178] Figure 6 The waveforms represent the dynamic speed control response of the motor as it jumps from 500 r / min to 2500 r / min, with time on the horizontal axis. The diagrams simultaneously plot the traditional ZigBee communication closed-loop PID speed curve, the open-loop speed curve of this invention's gain scheduling, and the real-time curve of the normalized frequency offset. Traditional solutions suffer from a 20ms communication delay, a 185ms speed response time, large overshoot, and susceptibility to electromagnetic interference. This invention eliminates secondary-side communication feedback, achieving 90% of the target speed in just 82ms, a response speed improvement of approximately 2.3 times, and a steady-state speed error of less than 1.8%. The corresponding frequency offset is smoothed by a 50ms time constant first-order low-pass filter, eliminating high-frequency oscillations. All time delays, speed accuracy, and dynamic indicators are measured values, clearly demonstrating the faster and more stable speed control performance of this solution.

[0179] Example

[0180] This embodiment uses a wireless-driven permanent magnet synchronous motor (PMSM) system with a rated power of 500 watts and a nominal resonant frequency of 85 kHz as a prototype verification platform to describe in detail the complete implementation process and experimental verification results of the present invention.

[0181] (1) System hardware platform configuration

[0182] Primary-side power stage: Employs a full-bridge inverter topology, using Cree C3M0065090D silicon carbide MOSFETs (rated withstand voltage 900 volts, on-resistance 65 milliohms) as power devices. The bus voltage is 48 volts, and the bus support capacitors are a combination of metal film capacitors with low equivalent series resistance. The primary-side compensation network uses an SS series compensation topology, with compensation inductors... (Litz wire wound, Q value greater than 200), digitally adjustable capacitor array It is implemented using an 8-bit digitally controlled binary weighted capacitor network, with a capacitance adjustment range of 47 to 68 nanofarads and an adjustment step resolution of less than 0.1 nanofarads; the magnetic coupling mechanism consists of a pair of circular planar helical coils with a diameter of 200 mm and a rated air gap of 50 mm. The coupling coefficient is within the range of coil offset ±30 mm and angular deviation ±15 degrees. The variation range is from 0.18 to 0.42; the secondary side compensation network parameters are... , and The design is symmetrical; the secondary side uses an uncontrolled full-bridge rectifier with DC bus capacitor filtering; the load motor is a permanent magnet synchronous motor with a rated power of 400 watts, a rated speed of 3000 rpm, and 4 pole pairs. The main control chip is the TI TMS320F28379D dual-core digital signal controller, where CPU1 (200 MHz C28x core) is responsible for executing the real-time control algorithm calculations of steps S100 to S300, and CPU2 (200 MHz CLA coprocessor) is responsible for executing the dead time calculation of step S400 and the safety protection monitoring of step S500.

[0183] (2) System initialization and soft start verification (step S100)

[0184] After the system powers on, CPU1 first sends a WHO_AM_I command to the digitally adjustable capacitor driver chip (Maxim Integrated MAX14745) at both the transmitter and receiver via an isolated SPI bus (10 megabits per second transmission rate). Within 78 microseconds, it receives a correct ID response, and the link self-test passes. CPU1 then reads the nominal resonant frequency from the on-chip FLASH. The capacitor array code value was initialized to 128 (corresponding to an intermediate capacitor value of approximately 57.5 nanofarads). The PWM duty cycle increased linearly from 5% at a rate of 1% per millisecond, reaching the rated 45% duty cycle after 40 milliseconds. During the ramp-up process, the peak bus current was effectively limited to below 2.1 amps (the hardware overcurrent threshold is 5 amps), and no overcurrent protection was triggered.

[0185] (3) Verification of resonant frequency tracking performance (step S200)

[0186] After stable operation, the DSP superimposes a linear frequency modulated (LFM) perturbation signal ranging from 72.25 kHz to 97.75 kHz onto the power carrier at a period of 10 milliseconds (the perturbation amplitude is controlled at 3% of the power carrier amplitude, corresponding to an additional power of less than 0.1% of the rated power). The Goertzel algorithm achieves a frequency resolution of 100 Hz. The total time for extracting amplitude and phase data at each frequency point, reconstructing the impedance frequency curve, and searching for poles is less than 500 microseconds, and is completed entirely within a single control cycle (100 microseconds) (frequency sweeping and data processing are background tasks, and the results are updated at the end of the cycle). The tracking results under standard 50 mm air gap and different load conditions are shown in Table 1.

[0187] Table 1 Frequency tracking results under different loads

[0188] .

[0189] (4) Verification of gain scheduling speed regulation performance (step S300)

[0190] During the factory calibration phase, offline experiments were conducted in a two-dimensional working domain with a coverage load rate of 10% to 120% and a coupling coefficient of 0.18 to 0.42 to establish... The range is to (0.1% increment) The mapping table has 80 lookup nodes. The monotonicity of the gain scheduling function within the effective operating range is determined using the Spearman rank correlation coefficient. (Statistical significance) Statistical validation was obtained. The user set the target speed to 1500 revolutions per minute (corresponding to the target armature voltage). The back electromotive force constant Pressure drop compensation The DSP obtains the result through a binary search in the gain scheduling table. Target frequency .like Figure 7 The actual speed of the motor shown smoothly increased from the current speed to 1503 rpm within 82 milliseconds (steady-state error +0.2%, far better than the design target of less than 2%). Under the subsequent applied step disturbance of 0 to 80% of rated load, the maximum transient speed fluctuation was only ±4.2%, and it recovered to steady state within 150 milliseconds.

[0191] (5) Dynamic dead zone and ZVS performance verification (step S400)

[0192] The datasheet for the silicon carbide MOSFET C3M0065090D specifies the nominal turn-off delay. Considering the difference in propagation delay of the driving circuit, set The proportionality coefficient was obtained by measuring the boundary dead time of the ZVS under different load current conditions and performing linear regression. (goodness of fit) During the entire process of the load stepping from 20% to 100%, the DSP linearly adjusted the dead time from approximately 54 nanoseconds to approximately 91 nanoseconds in real time. (Bridge arm midpoint voltage) The waveform is as follows Figure 8 The display shows the gate drive signal of the switch transistor. Before the rising edge arrives, The resonant transition has been completed to below 1.8 volts (less than 4% of the bus voltage). The ZVS conditions were consistently met under all test conditions. The inverter's measured efficiency ranged from 93.1% to 96.8% across the full load range, with an average efficiency of 95.4%.

[0193] (6) Multimodal security protection verification (step S500)

[0194] To verify the multi-mode protection function of step S500, an aluminum foreign object measuring 20 mm × 20 mm × 2 mm was suddenly inserted into the coil gap at a distance of 50 mm while the system was operating stably at 86.3 kHz and with an output power of 420 W. The eddy current effect of the foreign object caused the actual efficiency to deviate rapidly from the normal value: efficiency residual. The degradation worsened from -2% to -18% within approximately 1.2 milliseconds and continued to deteriorate. The DSP detected this over four consecutive cycles (400 microseconds). Then, the power reduction protection mode is triggered, shifting the operating frequency from 86.3 kHz to [a lower frequency] within 500 microseconds. ( (Offset +6.6%). At this safe frequency, the measured voltage gain dropped to 24% of the normal value, the secondary output voltage plummeted from 48 volts to 7.8 volts, and the motor output torque dropped to approximately 6% of the rated value (just enough to overcome static friction and maintain low-speed rotation). After troubleshooting, the system recovered to normal frequency and full power output within 3 control cycles (300 microseconds) without requiring a restart. Figure 9 The waveform diagrams from the multimodal safety protection experiment illustrate the dynamic changes in efficiency residual, operating frequency, and secondary-side output voltage after the intrusion of a metallic foreign object. As shown in the figure, under stable operating conditions of 86.3kHz and 420W output power, after the insertion of an aluminum foreign object, the efficiency residual rapidly deteriorated from -2% to -18% within approximately 1.2ms. After detecting the anomaly for four consecutive control cycles, power reduction protection was triggered. Subsequently, the controller actively shifted the operating frequency to a safe frequency of approximately 92.0kHz within approximately 500µs, causing the secondary-side output voltage to drop from 48V to 7.8V, achieving physical layer suppression of energy transfer. After fault resolution, the system recovered to its normal operating frequency and rated output state within approximately 300µs, verifying the rapid identification, active power reduction, and rapid recovery capabilities of this method under conditions without secondary-side communication.

[0195] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A wireless drive motor variable frequency speed control method based on adaptive resonant frequency tracking and gain scheduling, characterized in that, In a magnetically coupled resonant wireless power drive motor system, a wideband sweep frequency perturbation is superimposed on the power carrier to extract the amplitude and phase information of the primary voltage and current, reconstruct the impedance frequency response curve, and update the operating frequency at the current moment through minimum value search and filtering. Then, the DC bus voltage after rectification and filtering of the secondary side is obtained based on the normalized frequency offset. The target armature DC bus voltage required is calculated according to the target speed command. The target operating frequency of the primary inverter is calculated by reverse table lookup using the gain scheduling function.

2. The wireless drive motor variable frequency speed control method based on adaptive resonant frequency tracking and gain scheduling according to claim 1, characterized in that, The specific steps to obtain the current operating frequency include: (S210) Injection of broadband perturbation signal A set of linear frequency modulated (FM) sweep perturbation signals is superimposed on a normal power transmission sinusoidal carrier wave, with a sweep range of [missing information]. The amplitude of the perturbation signal Power carrier amplitude Within 5%; (S220) Carrier center alignment synchronization amplitude and phase acquisition At each perturbation frequency point At this point, the dual-channel ADC is triggered to synchronously acquire the instantaneous voltage value across the primary resonant circuit. and instantaneous current value The sampling window length corresponds to at least one complete perturbation signal period; for the acquired time-domain data, the Goertzel algorithm is used to perform single-frequency discrete Fourier transform calculations at the frequency points. Calculate the voltage amplitude at the location Current amplitude and the phase difference between voltage and current ; (S230) Reconstruction of the frequency response of the primary-side equivalent input impedance Based on collection The amplitude and phase data at each discrete frequency sweep point are used to reconstruct the primary-side equivalent input impedance in real time according to the impedance definition. Discrete frequency response curves within the swept frequency range: (S240) Resonant pole localization and precise sub-frequency resolution localization On the discrete impedance frequency curve reconstructed in step (S230), search for the impedance magnitude. The minimum point is found, and the sub-frequency resolution is accurately located between the found discrete minimum point and its two adjacent frequency points on the left and right. (S250) First-order low-pass filter adaptive frequency update The original optimal resonant frequency identified in step (S240) After smoothing, update to the system's current operating frequency: in, This is the operating frequency setpoint output to the PWM module for the current control cycle. The original optimal resonant frequency identified in the current frequency sweep cycle. This is the operating frequency of the previous control cycle. These are the filter coefficients.

3. The wireless drive motor variable frequency speed control method based on adaptive resonant frequency tracking and gain scheduling according to claim 2, characterized in that, The specific steps to obtain the target operating frequency of the primary-side inverter include: (S310) Calculation of normalized frequency offset The normalized frequency offset is calculated using the following formula: in, The nominal resonant frequency specified at the factory. The optimal resonant frequency is tracked in real time during the current control cycle via step (S250); (S320) Offline calibration and online table lookup of gain scheduling function Based on the gain scheduling function covering the entire operating range, an online table is looked up using the normalized frequency offset. The DC bus voltage after secondary-side rectification and filtering is obtained through calculation or function approximation. ; (S330) Target armature voltage feedforward calculation According to the set target speed command The required target armature DC bus voltage is calculated using the back electromotive force equation of the permanent magnet synchronous motor: in, Let be the back electromotive force constant of the motor. For the target mechanical angular velocity, To the current effective value of the load current Related comprehensive pressure drop compensation items; (S340) Inverse lookup table of gain scheduling function and target operating frequency output The target armature DC bus voltage calculated according to step (S330) In the gain scheduling function, a reverse lookup table operation is performed to find the target normalized frequency offset. ; Calculate the target operating frequency of the primary-side inverter based on the search results. .

4. The wireless drive motor variable frequency speed control method based on adaptive resonant frequency tracking and gain scheduling according to claim 3, characterized in that, It also includes system initialization and secure soft boot, the specific process of which is as follows: (S110) End-to-end hardware authentication and communication self-test The controller sends a query command to the driver chip of the digitally adjustable capacitor at the transmitting and receiving ends. If the controller receives a matching ID response and there is no fault status in the status register within the preset time window, it is determined that the full-link communication is normal and allows the subsequent startup process to proceed. If the self-test fails, the controller will forcibly pull down all 6 PWM outputs within a nanosecond time. (S120) Neutral impedance safety parameter preset The controller reads the nominal resonant frequency written during the factory calibration phase. The period register of the PWM timer of the primary inverter is initialized to the corresponding value, and the digital adjustable capacitor array at the transmitting end is also initialized. and receiver digitally adjustable capacitor array All are set to their middle code values; (S130) Current closed-loop linear ramp soft start Start the PWM module, set the initial duty cycle to a preset safe low threshold of 5%, and execute the current closed-loop ramp-up algorithm.

5. The wireless drive motor variable frequency speed control method based on adaptive resonant frequency tracking and gain scheduling according to claim 4, characterized in that, It also includes dynamic dead-zone adjustment and global soft switching, the specific process of which is as follows: (S410) Real-time detection of the root mean square value of the load current A current sampling resistor is connected in series between the source of the lower arm MOSFET and power ground of the primary side inverter bridge arm. The differential voltage signal across the resistor is amplified by an instrumentation amplifier with a fixed gain, and then sent to the controller through an anti-aliasing low-pass filter; continuous sampling is performed at a sampling rate synchronized with the control cycle; the controller samples the instantaneous current sequence in each control cycle. Perform root mean square calculation of the current effective value of the load current. , This represents the number of sampling points within a single period. (S420) Dead-time adaptive feedforward calculation Based on the current effective value of the load current Based on the switching characteristic parameters, the optimal dead time required for the current operating condition is calculated in real time according to the feedforward formula: ; The absolute minimum dead time to ensure safe straight-through of the bridge arm; This refers to the dead zone-current sensitivity coefficient. (S430) Dead time and working frequency are coordinated for atomic updates The dead time calculated in step (S420) Write the high-resolution dead-time configuration register of the PWM timer module to set the target operating frequency determined in step (S340). After being converted into a PWM timer period value, it is written into the corresponding register; (S440) Real-time monitoring and closed-loop compensation of zero-voltage switching status Real-time monitoring of the voltage at the midpoint of the primary-side inverter bridge arm and the gate drive signal of the switching transistor. The voltage level just before the rising edge arrives; If the voltage at the midpoint of the bridge arm If the system has naturally transitioned to near-zero level through LC resonance, it is determined that the current ZVS condition is met and the system is operating normally. The controller collects the ZVS status flag in each control cycle. If the ZVS status is lost for three consecutive control cycles, the dead time compensation fine-tuning mechanism is automatically triggered.

6. The wireless drive motor variable frequency speed control method based on adaptive resonant frequency tracking and gain scheduling according to claim 5, characterized in that, It also includes multimodal security protection based on frequency active backoff, specifically including: (S510) Real-time online calculation of efficiency residuals Define the efficiency residual of the current control cycle as: in, The actual input power for the current control cycle. This represents the theoretical expected load power of the secondary side under the current operating conditions. This is the baseline value for expected efficiency. (S520) Three-level threshold classification judgment for abnormal operating conditions Level 1, Warning Threshold :when First time below At this time, the system records abnormal event logs and increases the efficiency residual monitoring frequency from once per control cycle to multiple sliding window average calculations within each control cycle, but the system maintains normal operation and does not trigger any power limiting measures. Level 2, power reduction threshold :when consistently below Furthermore, if this state is maintained for more than 3 complete control cycles, the system determines that there is a non-transient, continuous abnormal energy loss, automatically triggers the power reduction protection mode, and enters the frequency active withdrawal process in sub-step S530. Level 3, Emergency Stop Threshold :when Fall to When the bus current hardware overcurrent comparator is triggered, the system immediately performs a hardware-level emergency shutdown, directly blocking all 6 PWM output signals within nanoseconds. (S530) Frequency active pullback physical layer energy suppression When sub-step S520 determines that the power reduction protection mode has been triggered, the controller executes the frequency active pullback strategy, reducing the inverter operating frequency from the current resonant frequency. Quickly shift to a predefined safe frequency .

Images