Optimal dynamic response deadbeat current prediction system for permanent magnet synchronous motor

Abstract

This invention relates to the field of motor control technology and discloses an optimal dynamic response deadbeat current prediction system for permanent magnet synchronous motors. The system performs the following steps: acquiring motor operating status data and historical sequence data; calculating the error phase trajectory and detecting periodic disturbance patterns through cross-mechanical cycle normalized correlation analysis; extracting historical feedforward compensation voltage vectors and calculating the net available voltage amplitude under the periodic disturbance pattern; calculating the upper limit of the current change rate based on the net available voltage amplitude and planning the optimal intermediate current trajectory; performing two-step deadbeat current prediction calculations to calculate the nominal voltage vector; synthesizing the voltage vector and performing constraint verification; outputting voltage commands and updating the angle index memory table.

Description

Technical Field

[0001] This invention relates to the field of motor control technology, and more specifically, to an optimal dynamic response deadbeat current prediction system for permanent magnet synchronous motors. Background Technology

[0002] In driving scenarios such as air conditioning compressors in electric vehicles, permanent magnet synchronous motors face two types of operating constraints: On the one hand, the load torque fluctuations caused by the compressor's intake and exhaust processes are related to the rotor's mechanical angle, exhibiting a periodic characteristic that repeats in each mechanical cycle, resulting in the dq axis current tracking error exhibiting a repetitive pattern with the mechanical angle; on the other hand, the DC bus voltage fluctuates within a wide range depending on the state of charge of the power battery and instantaneous power demand, and instantaneous voltage drops occur during rapid acceleration or when multiple electrical devices operate in parallel, causing the maximum voltage vector amplitude that the inverter can output to change in real time.

[0003] Existing deadbeat current prediction methods calculate the current change rate constraint based on the current bus voltage and plan the optimal intermediate current trajectory sequence to compensate for the prediction deviation caused by load disturbances in a pure feedback manner.

[0004] The existing technology has the following drawbacks: When periodic load disturbances and bus voltage drops occur concurrently, the pure feedback prediction error compensation requires a delay in error generation, detection, and compensation for the same disturbance pattern that recurs in each mechanical cycle, and cannot suppress periodic disturbances of known patterns in advance; if feedforward prediction based on historical compensation experience is introduced, the stored historical feedforward compensation voltage values ​​are recorded under previous bus voltage conditions. When the bus voltage drops, the historical feedforward compensation voltage values ​​may exceed the current voltage constraint capability, resulting in incomplete compensation or even voltage saturation after being truncated by the inverter; in addition, the planning of the optimal intermediate current trajectory sequence does not include the occupancy of the feedforward compensation voltage in the voltage budget when tightening the voltage constraint. The planned upper limit of the current change rate is actually unexecutable after the feedforward voltage is superimposed, causing the predicted voltage to exceed the inverter output capability, and the optimal dynamic response trajectory cannot be fully realized. Summary of the Invention

[0005] This invention provides an optimal dynamic response deadbeat current prediction system for permanent magnet synchronous motors, solving the technical problems in related technologies such as the lack of adaptive identification mechanism in current control of permanent magnet synchronous motors when facing periodic and non-periodic disturbances, the difficulty in balancing dynamic response and voltage constraints in deadbeat predictive control, and the lack of coordinated optimization between feedforward compensation and trajectory planning.

[0006] This invention discloses an optimal dynamic response deadbeat current prediction method for permanent magnet synchronous motors, including: acquiring the measured value of dq axis current, the reference value of dq axis current, the rotor mechanical angle, the rotor electrical angular velocity, the real-time DC bus voltage value at the current sampling moment, as well as the historical sequence of dq axis current tracking error and the remaining intermediate current trajectory sequence stored in the previous control cycle within the sliding window; Based on the historical sequence of the dq axis current tracking error, the error value is normalized and mapped according to the rotor mechanical angle. The normalized correlation coefficient between the trajectory point sequence of the current mechanical cycle and the corresponding angle position of the previous complete mechanical cycle is calculated. According to the comparison result of the normalized correlation coefficient and the periodicity determination threshold, the current state is marked as a periodic disturbance mode or a non-periodic disturbance mode. In periodic disturbance mode, the historical feedforward compensation voltage vector is extracted from the angle index memory table based on the predicted mechanical angle, and its magnitude is subtracted from the current maximum available voltage vector magnitude to obtain the net available voltage magnitude; in non-periodic disturbance mode, the current maximum available voltage vector magnitude is directly used as the net available voltage magnitude. The upper limit of the dq axis current change rate is calculated based on the net available voltage amplitude. Based on the comparison between the relative change of the upper limit of the change rate and the replanning trigger threshold, it is decided to extract the intermediate target current value from the remaining intermediate current trajectory sequence or to perform a cycle-by-cycle recursive calculation to generate a new optimal intermediate current trajectory sequence and extract the intermediate target current value. Using the intermediate target current value as the objective, a two-step deadbeat prediction operation is performed based on the discretized current prediction model of permanent magnet synchronous motor to calculate the nominal voltage vector; In periodic disturbance mode, the nominal voltage vector is superimposed with the historical feedforward compensation voltage vector to generate a composite voltage vector. After performing amplitude constraint verification on the composite voltage vector, the final dq-axis voltage command value is generated. In non-periodic disturbance mode, the nominal voltage vector is performed to generate the final dq-axis voltage command value. The final dq axis voltage command value is output to the inverter after space vector modulation, and the angle index memory table is updated with the actual feedforward compensation voltage value in periodic disturbance mode.

[0007] Furthermore, the normalized correlation coefficient is calculated using the Pearson correlation coefficient formula, specifically as follows: the error values ​​at each angle grid point in the error sequence after angle alignment within the current machine cycle and the error sequence at the corresponding angle position within the previous complete machine cycle are subtracted from the mean of their respective sequences, and then multiplied point by point and summed. The sum is divided by the product of the square roots of the sum of the squares of the mean deviations of the two sequences, and the quotient is the normalized correlation coefficient. When the normalized correlation coefficient exceeds the periodicity determination threshold and the number of consecutive mechanical cycles that meet this condition reaches the number of confirmed cycles, the current state is marked as a periodic disturbance mode.

[0008] Furthermore, the normalization mapping refers to rearranging the error values ​​at each sampling time from the time domain to the mechanical angle domain using the corresponding rotor mechanical angle as an index; When the rotational speed changes, resulting in different numbers of sampling points within the same mechanical angle range, a linear interpolation algorithm is used to resample the trajectory point sequences of the two cycles in the angle domain, so that the trajectory point sequences are aligned on the same angle grid.

[0009] Furthermore, the predicted mechanical angle is the current rotor mechanical angle plus the rotor electrical angular velocity divided by the number of motor pole pairs and then multiplied by the control cycle duration to obtain the angle increment. The angle index memory table uses the rotor mechanical angle as the index to uniformly discretize a complete mechanical rotation cycle into multiple angle partitions. Each angle partition stores the d-axis and q-axis components of the feedforward compensation voltage that was most recently executed in that partition. When the amplitude of the historical feedforward compensation voltage vector exceeds the product of the current maximum available voltage vector amplitude and the upper limit of the occupancy ratio, the historical feedforward compensation voltage vector is scaled proportionally according to the upper limit of the occupancy ratio, and the scaled amplitude is deducted from the current maximum available voltage vector amplitude as the feedforward occupancy amount.

[0010] Furthermore, the calculation process for the upper limit of the dq axis current change rate is as follows: based on the net available voltage amplitude, the amplitude of the back electromotive force voltage component determined by the rotor electric angular velocity, dq axis inductance, the current measured value of dq axis current and the permanent magnet flux linkage is subtracted to obtain the remaining available voltage amplitude. The upper limit of the dq-axis current change rate is obtained by dividing the product of the remaining available voltage amplitude and the control cycle duration by the square root of the sum of the squares of the d-axis inductance and the q-axis inductance. The relative change of the upper limit of the change rate is the ratio of the absolute value of the difference between the current upper limit of the change rate and the upper limit of the change rate of the previous control cycle to the upper limit of the change rate of the previous control cycle.

[0011] Furthermore, when the relative change of the upper limit of the rate of change exceeds the replanning trigger threshold, starting from the current measured value of the dq-axis current, taking the updated upper limit of the rate of change of the dq-axis current as the constraint and the reference value of the dq-axis current as the terminal target, a new optimal intermediate current trajectory sequence is generated by recursive calculation on a cycle-by-cycle basis. The first target value is extracted from the new optimal intermediate current trajectory sequence, and the deviation between the first target value and the target value at the corresponding time in the old remaining intermediate current trajectory sequence is subjected to amplitude limiting and smoothing processing to obtain the smoothed intermediate target current value. The aforementioned amplitude limiting and smoothing process refers to truncating the deviation to the smoothing amplitude limit when the absolute value of the deviation exceeds the smoothing amplitude limit, and using the old target value plus the truncated deviation value as the intermediate target current value. When the absolute value of the deviation does not exceed the smoothing amplitude limit, the initial target value is directly used as the intermediate target current value. When the remaining intermediate current trajectory sequence is empty and the relative change of the upper limit of the rate of change does not exceed the replanning trigger threshold, the cycle-by-cycle recursive calculation is also performed to generate a new optimal intermediate current trajectory sequence and the first target value is extracted as the intermediate target current value.

[0012] Furthermore, the process of the two-step deadbeat prediction operation is as follows: Based on the forward Euler discretization difference equation of the dq-axis voltage equation of the permanent magnet synchronous motor, the dq-axis current value after the first step is predicted using the current measured dq-axis current value as the initial value and the nominal voltage vector to be determined. Then, the dq-axis current value after the first step is predicted using the predicted dq-axis current value as the initial value. The predicted dq-axis current value after the second step is set to be equal to the intermediate target current value. The d-axis and q-axis components of the nominal voltage vector to be applied are then calculated in reverse.

[0013] Furthermore, the amplitude constraint verification process is as follows: When the magnitude of the synthesized voltage vector exceeds the current maximum output voltage vector magnitude, the scaling factor is calculated as the ratio of the current maximum output voltage vector magnitude to the synthesized voltage vector magnitude. The nominal voltage vector and the historical feedforward compensation voltage vector are scaled proportionally by the scaling factor and then added together to generate the final dq axis voltage command value. When the scaling factor is lower than the minimum scaling threshold, the amplitude of the historical feedforward compensation voltage vector is reduced by a preset reduction ratio and then superimposed with the nominal voltage vector to generate the final dq axis voltage command value. The reduced historical feedforward compensation voltage vector is used as the feedforward value actually used in the current control cycle for updating the angle index memory table.

[0014] Furthermore, the update of the angle index memory table adopts a weighted average method, which is to write back the corresponding angle partition after weighting and fusing the feedforward compensation voltage value actually executed in the current control cycle with the original stored value in the angle index memory table according to a fixed weight coefficient; When the current state switches from aperiodic disturbance mode to periodic disturbance mode, or when the change in DC bus voltage between adjacent control cycles exceeds the voltage jump threshold, the values ​​of all storage units in the angle index memory table are attenuated according to a preset attenuation coefficient.

[0015] This invention provides an optimal dynamic response deadbeat current prediction system for permanent magnet synchronous motors, comprising: The data acquisition module is used to acquire the measured value of the dq axis current, the reference value of the dq axis current, the rotor mechanical angle, the rotor electrical angular velocity, the real-time DC bus voltage value, the historical sequence of the dq axis current tracking error within the sliding window, and the remaining intermediate current trajectory sequence stored in the previous control cycle. The periodic disturbance detection module is used to normalize the error value according to the rotor mechanical angle based on the historical sequence of the dq axis current tracking error, calculate the normalized correlation coefficient between the trajectory point sequence of the corresponding angle position of the current mechanical cycle and the previous complete mechanical cycle, and mark the current state as a periodic disturbance mode or a non-periodic disturbance mode according to the comparison result of the normalized correlation coefficient and the periodicity determination threshold. The net available voltage calculation module is used to extract the historical feedforward compensation voltage vector from the angle index memory table based on the predicted mechanical angle under periodic disturbance mode and subtract its magnitude from the current maximum available voltage vector magnitude to obtain the net available voltage magnitude. Under non-periodic disturbance mode, the current maximum available voltage vector magnitude is directly used as the net available voltage magnitude. The trajectory planning module is used to calculate the upper limit of the dq axis current change rate based on the net available voltage amplitude, and extract the intermediate target current value from the remaining intermediate current trajectory sequence or perform cycle-by-cycle recursive calculation to generate a new optimal intermediate current trajectory sequence and extract the intermediate target current value based on the comparison result of the relative change of the upper limit of the change rate and the replanning trigger threshold. The deadbeat prediction module is used to perform two-step deadbeat prediction operations based on the discretized current prediction model of permanent magnet synchronous motor with the intermediate target current value as the target, and calculate the nominal voltage vector. The voltage synthesis and constraint verification module is used to superimpose the nominal voltage vector and the historical feedforward compensation voltage vector to generate a synthesized voltage vector under periodic disturbance mode, and then perform amplitude constraint verification to generate the final dq axis voltage command value. Under non-periodic disturbance mode, the module performs amplitude constraint verification on the nominal voltage vector to generate the final dq axis voltage command value. The output and update module is used to output the final dq axis voltage command value to the inverter after space vector modulation, and update the angle index memory table with the actual feedforward compensation voltage value executed in the periodic disturbance mode.

[0016] This invention detects periodic disturbance patterns by performing cross-mechanical cycle normalized correlation analysis on the dq-axis error phase trajectory, solving the technical problem that pure feedback deadbeat prediction cannot suppress known pattern periodic disturbances in advance. It achieves the technical effect of distinguishing stable repetitive error patterns caused by periodic load disturbances from random error fluctuations in transient transition processes, thus avoiding misjudgments. By using the predicted mechanical angle as an index to extract historical feedforward compensation voltage vectors from the angle index memory table in advance, it eliminates the inherent response delay of pure feedback deadbeat prediction to known pattern disturbances, enabling predictive suppression of periodic components in current tracking errors. Finally, by subtracting the amplitude of the historical feedforward compensation voltage vector from the current maximum available voltage vector amplitude, the net available voltage is obtained. The system calculates the voltage amplitude value and plans the optimal intermediate current trajectory sequence under the constraint of net available voltage amplitude, solving the technical problem that the synthesized voltage vector exceeds the inverter output capacity due to the failure to include the feedforward compensation voltage occupancy in the voltage budget. By updating the angle index memory table with the actual executed feedforward compensation voltage value, the data stored in the angle index memory table adaptively reflects the achievable compensation level under the current voltage constraint condition, forming a closed-loop adaptive adjustment. By triggering the replanning of the optimal intermediate current trajectory sequence when the voltage constraint changes significantly and applying amplitude limiting and smoothing processing to the switching between the old and new trajectories, the current tracks the reference value along the fastest feasible path allowed by the current constraint condition, achieving the best dynamic response under the concurrent conditions of dynamic changes in voltage constraint and periodic load disturbance. Attached Figure Description

[0017] Figure 1 This is a flowchart of the optimal dynamic response deadbeat current prediction method for permanent magnet synchronous motors provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the comparison of the dq axis current error angle distribution (current cycle vs. previous cycle) provided in an embodiment of the present invention. Figure 3 This is a schematic diagram comparing the voltage amplitudes at each step provided in the embodiments of the present invention; Figure 4 This is a schematic diagram of the dq-axis current state change provided by an embodiment of the present invention: measured value → intermediate target → reference value; Figure 5 This is a schematic diagram of the dq-axis voltage vector synthesis decomposition (d-axis and q-axis components) provided in an embodiment of the present invention; Figure 6 This is a schematic diagram comparing the feedforward voltage before and after updating the angle index memory table according to an embodiment of the present invention; Figure 7 This is a schematic diagram comparing the upper limit of the current change rate with the trigger replanning threshold provided in an embodiment of the present invention; Figure 8 This is a schematic diagram illustrating the relationship between bus voltage drop and maximum available voltage provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of the dq-axis current prediction error analysis in the two-step deadbeat prediction provided in an embodiment of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be described in detail below with reference to the accompanying drawings. The description in this part is only exemplary and explanatory, and should not be used to limit the scope of protection of the present invention in any way.

[0019] Deadbeat-free current predictive control of permanent magnet synchronous motors predicts the required voltage vector based on the motor's discretized current model in each control cycle, enabling the actual current to accurately track the reference value within a finite number of steps, thus exhibiting rapid dynamic response capabilities. In drive scenarios such as air conditioning compressors in electric vehicles, permanent magnet synchronous motors face two types of operating constraints: Firstly, the load torque fluctuations caused by the compressor's intake and exhaust processes are related to the rotor's mechanical angle, exhibiting a periodic characteristic that repeats in each mechanical cycle, resulting in a repetitive pattern of dq-axis current tracking error with respect to the mechanical angle; secondly, the DC bus voltage fluctuates within a wide range depending on the state of charge of the power battery and instantaneous power demand, and instantaneous voltage drops occur during rapid acceleration or when multiple electrical devices operate in parallel, causing the maximum output voltage vector amplitude of the inverter to change in real time.

[0020] Existing deadbeat current prediction methods calculate current change rate constraints based on the current bus voltage and plan the optimal intermediate current trajectory sequence, using pure feedback to compensate for prediction deviations caused by load disturbances. When periodic load disturbances and bus voltage drops occur concurrently, the pure feedback prediction deviation compensation requires a delay in error generation, detection, and compensation for the same disturbance pattern that recurs in each mechanical cycle, making it impossible to suppress periodic disturbances of known patterns in advance. If feedforward prediction based on historical compensation experience is introduced, the stored historical feedforward compensation voltage values ​​are recorded under previous bus voltage conditions. When the bus voltage drops, the historical feedforward compensation voltage values ​​may exceed the current voltage constraint capability, resulting in incomplete compensation or even voltage saturation after being truncated by the inverter. Furthermore, the planning of the optimal intermediate current trajectory sequence does not include the occupancy of the feedforward compensation voltage in the voltage budget when tightening the voltage constraint. The planned upper limit of current change rate is actually unexecutable after the feedforward voltage is superimposed, causing the predicted voltage to exceed the inverter's output capability, and the optimal dynamic response trajectory cannot be fully realized.

[0021] According to an embodiment of this invention, a method for predicting the optimal dynamic response deadbeat current for permanent magnet synchronous motors is provided. It should be understood that the controller in this embodiment is a digital processor capable of performing data processing operations, installed in the permanent magnet synchronous motor drive system. It acquires the dq-axis current through a current sensor, obtains rotor position information through an encoder, acquires the DC bus voltage through a voltage sampling circuit, and outputs voltage commands to the inverter through space vector modulation.

[0022] At least one embodiment of the present invention discloses a method for predicting the best dynamic response deadbeat current for permanent magnet synchronous motors, such as... Figure 1 As shown, it includes the following steps: Step 1: Obtain motor operating status data and historical sequence data; At the current sampling moment, acquire the measured value of the dq-axis current, the reference value of the dq-axis current, the rotor mechanical angle, the rotor electrical angular velocity, and the real-time DC bus voltage. Simultaneously, read the most recent value within the sliding window. The historical sequence of dq-axis current tracking error at each sampling time and the remaining intermediate current trajectory sequence stored in the previous control cycle, among which This represents the number of sampling times within the sliding window.

[0023] It should be noted that the measured values ​​of the d-axis and q-axis currents mentioned above are in the synchronous rotating coordinate system after coordinate transformation. The reference values ​​for the d-axis and q-axis currents are the target d-axis and target q-axis currents output by the upper-level speed control loop or torque control loop. (The length of the sliding window is also mentioned.) The sampling points must cover at least one complete rotor mechanical rotation cycle to ensure the data integrity required for subsequent cross-cycle analysis.

[0024] It should be noted that the aforementioned remaining intermediate current trajectory sequence is a sequence of intermediate target current values ​​generated during the previous trajectory planning process but not yet executed, arranged in chronological order. When the controller starts up for the first time or when there is no valid remaining intermediate current trajectory sequence in the previous control cycle, the remaining intermediate current trajectory sequence is empty.

[0025] Step 2: Calculate the error phase trajectory and detect periodic disturbance patterns; Based on the measured and reference values ​​of the dq-axis current, the current dq-axis current error value is calculated, and the error change rate is obtained by differential operation on the dq-axis current error value. The current dq-axis current error value and the error change rate are used as coordinate components to determine the current phase trajectory point on the error phase plane.

[0026] The error values ​​at each sampling moment in the dq-axis current tracking error history sequence are normalized and mapped according to the corresponding rotor mechanical angle to obtain an error distribution sequence with the mechanical angle as the horizontal axis. The trajectory point sequence within the current mechanical cycle and the trajectory point sequence at the corresponding angular position within the previous complete mechanical cycle are extracted from the error distribution sequence. The normalized correlation coefficient between the two trajectory point sequences is calculated using the Pearson correlation coefficient formula. The normalized correlation coefficient is compared with a periodicity determination threshold. When the normalized correlation coefficient exceeds the periodicity determination threshold, the current state is marked as a periodic disturbance mode; when the normalized correlation coefficient does not exceed the periodicity determination threshold, the current state is marked as an aperiodic disturbance mode.

[0027] Furthermore, the above normalized correlation coefficient The calculation method is as follows: Let the error sequence after angle alignment within the current mechanical cycle be... The error sequence for the corresponding angular position within the previous complete mechanical cycle is as follows: ,in , This represents the number of angular grid points after alignment. and Sequences and The mean, For sequence The Middle Error values ​​at each angle grid point For sequence The Middle The error value at each angle grid point is:

[0028] The range of values ​​is , The closer The stronger the waveform repeatability between two sequences, the better. The periodicity threshold is... A positive threshold within the range of values, used to... near Strong periodic repetition state and Significant deviation Distinguish between non-periodic states.

[0029] It should be noted that the above-mentioned normalization mapping refers to rearranging the error values ​​at each sampling moment from the time domain to the mechanical angle domain. Specifically, it uses the rotor mechanical angle corresponding to each sampling moment as an index to rearrange the error values, allowing the error distribution under different speed conditions to be compared under a unified angular coordinate system. When the speed changes, the number of sampling points within the same mechanical angle interval changes. The normalization mapping uses a linear interpolation algorithm to resample the trajectory point sequences of the two cycles in the angle domain, aligning the trajectory point sequences on the same angular grid.

[0030] It should be noted that the value of the aforementioned periodicity determination threshold should be able to distinguish between the stable repetitive error pattern caused by periodic load torque and the random error fluctuations during the transient transition process.

[0031] In this embodiment, to avoid misjudging occasional error waveform similarities during transient transitions as periodic disturbance modes, after the normalized correlation coefficient exceeds the periodicity determination threshold, the number of consecutive mechanical cycles that meet this condition is counted. Only when the number of consecutively satisfied mechanical cycles reaches the confirmation cycle count is the current state marked as a periodic disturbance mode. By introducing a continuous confirmation condition, the determination of periodic disturbance modes becomes more robust.

[0032] Step 3: Extract the feedforward compensation voltage and calculate the net usable voltage amplitude; When the current state is marked as a periodic disturbance mode, the predicted mechanical angle one control cycle ahead is calculated based on the rotor electrical angular velocity and the current rotor mechanical angle. The corresponding historical feedforward compensation voltage vector is extracted from the angle index memory table using the predicted mechanical angle as the index. The magnitude of the historical feedforward compensation voltage vector is calculated, and the current maximum available voltage vector magnitude is calculated based on the real-time DC bus voltage value. The magnitude of the historical feedforward compensation voltage vector is subtracted from the current maximum available voltage vector magnitude to obtain the net available voltage magnitude after deducting feedforward occupancy.

[0033] When the current state is marked as an aperiodic disturbance mode, the current maximum available voltage vector magnitude is directly used as the net available voltage magnitude, and the historical feedforward compensation voltage vector extraction operation is not performed.

[0034] Furthermore, the above-mentioned predictive mechanical angle The calculation method is as follows:

[0035] in, The current rotor mechanical angle, The rotor's electric angular velocity, This represents the number of pole pairs of the motor. To control the cycle length, the predicted mechanical angle, rather than the current mechanical angle, is used as an index for extraction. This aligns the historical feedforward compensation voltage vector with the actual mechanical angle position at the moment of execution, compensating for the delay of one control cycle between sampling and execution.

[0036] It should be noted that the aforementioned angle index memory table is a lookup table indexed by the rotor's mechanical angle. The storage units of the angle index memory table correspond to the various angle partitions after a complete mechanical rotation cycle has been uniformly discretized. Each storage unit stores the d-axis and q-axis components of the most recently executed feedforward compensation voltage within that angle partition. During the controller initialization phase, the values ​​of each storage unit in the angle index memory table are set to zero, and are gradually updated and filled by step 7 during subsequent operation.

[0037] Furthermore, the aforementioned current maximum available voltage vector magnitude Based on real-time DC bus voltage The value is obtained by converting the linear voltage utilization factor of space vector modulation. In space vector modulation, there is a fixed proportional relationship between the maximum output voltage vector amplitude of the inverter and the DC bus voltage. This proportionality factor is determined by the modulation method of space vector modulation. For standard space vector modulation, this proportionality factor is... ,Right now .

[0038] In this embodiment, to prevent the amplitude of the historical feedforward compensation voltage vector from occupying an excessive proportion of the available voltage space when the bus voltage is low, thus resulting in insufficient nominal voltage margin for deadbeat prediction operations, an upper limit is set on the occupancy ratio of the historical feedforward compensation voltage vector. When the amplitude of the historical feedforward compensation voltage vector exceeds the product of the current maximum available voltage vector amplitude and the upper limit of the occupancy ratio, the historical feedforward compensation voltage vector is scaled proportionally according to the upper limit of the occupancy ratio, and the scaled amplitude is deducted from the current maximum available voltage vector amplitude as the feedforward occupancy amount. By setting an upper limit on the occupancy ratio, the net available voltage amplitude is always kept at a minimum, ensuring that deadbeat prediction operations have basic voltage operating space.

[0039] Step 4: Calculate the upper limit of the rate of change of current and plan the optimal intermediate current trajectory; Based on the net available voltage amplitude and rotor electric angular velocity, calculate the upper limit of the dq-axis current change rate at the current moment. Compare the upper limit of the dq-axis current change rate at the current moment with the upper limit of the change rate used in the trajectory planning of the previous control cycle to generate the relative change of the upper limit of the change rate.

[0040] Determine whether the relative change in the upper limit of the rate of change exceeds the replanning trigger threshold: If the upper limit of the rate of change does not exceed the replanning trigger threshold, it indicates that the current voltage constraint has not changed significantly relative to the previous control cycle. The intermediate target current value corresponding to the next control cycle is extracted from the remaining intermediate current trajectory sequence.

[0041] If the relative change of the upper limit of the rate of change exceeds the replanning trigger threshold, it indicates that the voltage constraint has changed significantly. Starting with the current measured value of the dq-axis current, the updated upper limit of the dq-axis current rate of change is the constraint, and the dq-axis current reference value is the terminal target, a new optimal intermediate current trajectory sequence is generated through a cycle-by-cycle recursive calculation. The first target value is extracted from the new optimal intermediate current trajectory sequence, and the deviation between the first target value and the target value at the corresponding time in the old remaining intermediate current trajectory sequence is subjected to amplitude limiting and smoothing to obtain the smoothed intermediate target current value.

[0042] Furthermore, the calculation of the upper limit of the dq-axis current change rate is based on the dq-axis voltage equation of the permanent magnet synchronous motor. Let... and These are the d-axis and q-axis inductances, respectively. For stator resistance, It is a permanent magnet flux linkage. The rotor's electric angular velocity, and These are the measured values ​​of the current along the d-axis and q-axis, respectively. The net usable voltage amplitude is the voltage amplitude remaining after deducting the back electromotive force voltage component, which can be used to drive the change in driving current. for:

[0043] d-axis and q-axis currents in a single control cycle Upper limit of rate of change within satisfy:

[0044] in, To control the cycle duration.

[0045] It should be noted that the above The dimension of is volt-second. The dimension of is Henry, the dimension of the ratio of the two is Ampere, and the upper limit of the rate of change is... The dimensions are consistent, and the dimensions of each term in the formula are matched.

[0046] Furthermore, the above Indicates a single control cycle The maximum allowable variation of the inner dq-axis current, i.e., based on the current measured value of the dq-axis current. Center of the circle Within a circular region with radius [radius value], the intermediate target current values ​​for the next control cycle all satisfy the voltage constraint. The cycle-by-cycle recursive calculation, in each step, along the direction towards the terminal target current value, [does not exceed] [the voltage constraint]. The magnitude of the advance ensures that the current increment in the generated optimal intermediate current trajectory sequence does not exceed a certain value in each control cycle. This ensures that the corresponding nominal voltage vector is within the net available voltage amplitude. It is executable within the constraints.

[0047] It should be noted that the relative change of the above-mentioned upper limit of the rate of change is the ratio of the absolute value of the difference between the current upper limit of the rate of change and the upper limit of the rate of change in the previous control period to the upper limit of the rate of change in the previous control period.

[0048] It should be noted that the above-mentioned cycle-by-cycle recursive calculation refers to: starting from the current measured current value, advancing along the direction towards the terminal target current value in each step with an amplitude not exceeding the upper limit of the rate of change, gradually generating intermediate target current values ​​for each control cycle until the trajectory reaches the terminal target value. The optimal intermediate current trajectory sequence generated by the cycle-by-cycle recursive calculation converges to the terminal target at the fastest speed within the rate of change constraint.

[0049] It should be noted that the above-mentioned limiting and smoothing process refers to: calculating the deviation between the first-step target value of the new optimal intermediate current trajectory sequence and the corresponding target value of the old remaining intermediate current trajectory sequence; when the absolute value of this deviation exceeds the smoothing limit, the deviation is truncated to the smoothing limit, and the smoothed intermediate target current value is obtained by adding the truncated deviation value to the corresponding target value of the old remaining intermediate current trajectory sequence. When the absolute value of the deviation does not exceed the smoothing limit, the first-step target value of the new optimal intermediate current trajectory sequence is directly used as the intermediate target current value.

[0050] In this embodiment, to adapt the replanning trigger threshold to the changing characteristics of voltage constraints under different operating conditions, the replanning trigger threshold is set in segments according to the current bus voltage range. When the bus voltage is in a lower range, a smaller replanning trigger threshold is used, making the trajectory planning more sensitive to changes in voltage constraints and tightening the constraints in time to avoid voltage saturation. When the bus voltage is in a higher range, a larger replanning trigger threshold is used to reduce unnecessary replanning times and lower computational overhead.

[0051] In this embodiment, the method further includes the following steps: when the remaining intermediate current trajectory sequence is empty and the relative change of the upper limit of the rate of change does not exceed the replanning trigger threshold, a new optimal intermediate current trajectory sequence is generated by recursive calculation on a cycle-by-cycle basis, and the first target value is extracted as the intermediate target current value. This step handles the situation at the initial stage of controller startup or after the optimal intermediate current trajectory sequence of the previous control cycle has been fully executed, ensuring that there is a valid intermediate target current value available for subsequent steps at any time.

[0052] Step 5: Perform two-step deadbeat current prediction calculation; The original dq-axis current reference value is replaced with the intermediate target current value. Based on the discretized current prediction model of permanent magnet synchronous motor, a two-step deadbeat prediction operation is performed to calculate the nominal voltage vector required for the dq-axis current to accurately reach the intermediate target current value after two control cycles.

[0053] Furthermore, the difference equation for the discretized current prediction model of the permanent magnet synchronous motor described above is:

[0054]

[0055] in, and For the first The d-axis and q-axis currents at time t. and For the first The d-axis and q-axis voltages applied at each moment, and For the first Predicted d-axis and q-axis current values ​​at time t. For discrete time indexes, To control the cycle duration, and These are the d-axis and q-axis inductances, respectively. For stator resistance, The rotor's electric angular velocity, This refers to the permanent magnet flux linkage. Two-step deadbeat prediction calculation means: first, based on the current measured dq-axis current and the voltage to be determined, predict the current value after the first step; then, using the predicted current value after the first step as the initial value, predict the current value after the second step; finally, set the predicted current value after the second step equal to the intermediate target current value, and use this to calculate the required applied voltage, which is the nominal voltage vector. The use of two-step prediction instead of one-step prediction is to compensate for the inherent control cycle calculation delay between sampling and actual voltage application in a digital control system.

[0056] Furthermore, the above reverse calculation process is as follows: Let the current time be... The current measured value of the dq axis current is and The intermediate target current value is and The nominal voltage vector to be determined is and .in, and These are the d-axis current value and the q-axis current value of the intermediate target, respectively. and These are the d-axis and q-axis components of the nominal voltage vector, respectively. First, let's... and Substituting into the difference equation, from and Calculate the first Current prediction at time and ; and then and Let the initial value be , and let the . The predicted current value at time t is equal to and The inverse solution of the difference equation yields the first... The voltage to be applied at a given time; due to the first... The voltage applied at time 1 Since the calculation needs to be completed immediately, the voltage obtained from the inverse solution above is used as... and Output. The electric angular velocity at both moments in the two-step prediction is taken as the value at the current moment. value.

[0057] Step 6: Synthesize the voltage vector and perform constraint verification; When the current state is marked as a periodic disturbance mode, the nominal voltage vector is vector-superimposed with the historical feedforward compensation voltage vector to generate a composite voltage vector. The current maximum output voltage vector amplitude is calculated based on the real-time DC bus voltage value, and amplitude constraint verification is performed on the composite voltage vector. If the amplitude of the composite voltage vector does not exceed the current maximum output voltage vector amplitude, the composite voltage vector is directly used as the final dq-axis voltage command value. If the amplitude of the composite voltage vector exceeds the current maximum output voltage vector amplitude, a scaling factor is calculated as the ratio of the current maximum output voltage vector amplitude to the composite voltage vector amplitude. This scaling factor is used to proportionally scale both the d-axis and q-axis components of the nominal voltage vector and the d-axis and q-axis components of the historical feedforward compensation voltage vector. The scaled nominal component is then added to the scaled feedforward component to generate the final dq-axis voltage command value.

[0058] When the current state is marked as a non-periodic disturbance mode, the nominal voltage vector is directly subjected to amplitude constraint verification and amplitude limiting processing based on the current maximum output voltage vector amplitude to generate the final dq axis voltage command value.

[0059] Furthermore, the amplitude constraint verification process for the above-mentioned synthesized voltage vector is as follows: Let the nominal voltage vector be... The historical feedforward compensation voltage vector is The synthesized voltage vector is ,in and These are the d-axis and q-axis components of the nominal voltage vector, respectively. and These are the d-axis and q-axis components of the historical feedforward compensation voltage vector, respectively. and These are the d-axis and q-axis components of the synthesized voltage vector, respectively. , The magnitude of the synthesized voltage vector is The current maximum output voltage vector amplitude is .when At that time, scaling factor ,in The scaling factor is used to determine the final dq axis voltage command value. This is equivalent to applying the same scaling factor to both the nominal component and the feedforward component simultaneously. .

[0060] It should be noted that the above method of scaling both the nominal and feedforward components proportionally ensures that both components shrink proportionally when the voltage saturates, maintaining the directional relationship between the nominal voltage and the historical feedforward compensation voltage in the synthesized voltage vector. Compared to methods that prioritize truncating the feedforward component or the nominal component, proportional scaling uniformly reduces the current tracking accuracy and periodic disturbance feedforward compensation capability of deadbeat prediction, avoiding the extreme situation where one function is completely sacrificed.

[0061] In this embodiment of the application, based on step 6, when the magnitude of the synthesized voltage vector exceeds the limit and the scaling factor... When the voltage falls below the minimum scaling threshold, it indicates that the current voltage constraint is extremely tight. The amplitude of the historical feedforward compensation voltage vector is reduced by a preset reduction ratio and then superimposed on the nominal voltage vector. The reduced historical feedforward compensation voltage vector is then passed to step 7 as the actual feedforward value used in the current control cycle to update the angle index memory table. By actively reducing the feedforward component under extreme voltage pressure conditions, more execution space is reserved for the nominal voltage, avoiding the loss of basic current tracking capability in deadbeat prediction operations due to excessive compression of the nominal voltage.

[0062] Furthermore, the aforementioned minimum scaling threshold is Range of values A positive threshold is used to identify the scaling factor. Extreme voltage stress state where the voltage is too small and the synthesized voltage vector is significantly compressed. When When the value is below the minimum scaling threshold, it indicates that scaling on a proportional basis is no longer sufficient to maintain feedforward compensation while preserving enough nominal voltage execution space. At this point, active reduction of the feedforward component is triggered.

[0063] Step 7: Output voltage command and update memory data; The final dq axis voltage command value is processed by space vector modulation and then output to the inverter, which drives the permanent magnet synchronous motor.

[0064] When the current state is marked as a periodic disturbance mode, the feedforward compensation voltage value actually executed in the current control cycle is written into the angle index memory table according to the angle partition index corresponding to the current rotor mechanical angle, and the feedforward compensation voltage value stored in the angle partition is updated.

[0065] The remaining intermediate current trajectory sequence after extracting the intermediate target current value in the current control cycle, along with the current upper limit of the rate of change, is stored in the buffer for use in the next control cycle.

[0066] It should be noted that the feedforward compensation voltage value actually executed above, when the synthesized voltage vector in step 6 does not exceed the limit, is the original historical feedforward compensation voltage vector extracted from the angle index memory table; when the synthesized voltage vector in step 6 exceeds the limit, it is the scaled historical feedforward compensation voltage vector. The angle index memory table is updated with the actual executed value rather than the original extracted value, so that the data stored in the angle index memory table reflects the actual achievable compensation level under the current voltage constraint. When the bus voltage remains at a low level, after several mechanical cycles of updating the angle index memory table, the feedforward compensation voltage values ​​stored in each angle partition naturally converge to an amplitude range matching the current voltage constraint. Therefore, the historical feedforward compensation voltage vector subsequently extracted from the angle index memory table will not excessively occupy voltage space when participating in the calculation of the net available voltage amplitude in step 3.

[0067] Furthermore, the update of the aforementioned angle index memory table adopts a weighted average method. The feedforward compensation voltage value actually executed in the current control cycle is weighted and fused with the original stored value in the angle index memory table according to a fixed weight coefficient before being written back to the corresponding angle partition. The weighted fusion ensures that the stored value in the angle index memory table converges smoothly during the update process of multiple consecutive machine cycles, avoiding sudden interference to the angle index memory table caused by a single abnormal execution value.

[0068] In this embodiment, to prevent the data in the angle index memory table from retaining conservative compensation values ​​from low-voltage conditions for an extended period after the bus voltage returns to normal, when the current state switches from aperiodic disturbance mode to periodic disturbance mode, or when the change in DC bus voltage exceeds the voltage jump threshold, the values ​​of all storage cells in the angle index memory table are attenuated using a preset attenuation coefficient. This attenuation process gradually reduces the compensation values ​​stored under the old conditions, accelerating the adaptation of the angle index memory table to the current operating conditions.

[0069] Furthermore, the aforementioned voltage jump threshold is a positive threshold expressed in voltage dimensions, used to determine whether the change in DC bus voltage between adjacent control cycles exceeds the normal fluctuation range. When the change in bus voltage exceeds the voltage jump threshold, a significant voltage condition switch is considered to have occurred, triggering the attenuation processing of the angle index memory table to accelerate the adaptation of the angle index memory table to the new voltage conditions.

[0070] The optimal dynamic response deadbeat current prediction method for permanent magnet synchronous motors provided in this embodiment distinguishes between stable repetitive error patterns caused by periodic load disturbances and random error fluctuations during transient transitions by performing cross-mechanical cycle normalized correlation analysis on the dq-axis error phase trajectory. This is because the normalized correlation coefficient... It directly measures the similarity of error trajectories in the angular domain within adjacent mechanical cycles, so when periodic disturbances exist... Approaching When in a transient transition process Significant deviation This avoids misjudging the periodic steady-state error trajectory as an oscillation and incorrectly reducing the compensation intensity.

[0071] Under periodic disturbance patterns, an angle index memory table is used to predict mechanical angles. A forward extraction is performed to align the historical feedforward compensation voltage vector with the mechanical angular position at the actual execution moment. Because periodic disturbances exhibit the same disturbance characteristics at the same angular position within each mechanical cycle, and the historical feedforward compensation voltage vector is acquired in advance at the sampling moment and participates in the voltage synthesis calculation, there is no need to wait for the error to occur before performing feedback compensation. This eliminates the inherent response delay of pure feedback deadbeat prediction to known pattern disturbances, and the periodic component in the current tracking error is suppressed by prediction.

[0072] Because the net available voltage amplitude is obtained by subtracting the amplitude of the historical feedforward compensation voltage vector from the current maximum available voltage vector amplitude in step 3, and the calculation of the upper limit of the dq axis current change rate and the planning of the optimal intermediate current trajectory sequence in step 4 are both performed under the constraint of the net available voltage amplitude, the nominal voltage vector required by the planned optimal intermediate current trajectory sequence and the composite voltage vector after superimposing the historical feedforward compensation voltage vector are within the output range of the inverter. This avoids the situation where the composite voltage vector exceeds the output capacity of the inverter due to the fact that the optimal intermediate current trajectory sequence planning does not consider the occupation of the historical feedforward compensation voltage vector.

[0073] Because replanning of the optimal intermediate current trajectory sequence is triggered when the upper limit of the rate of change exceeds the replanning trigger threshold, when the bus voltage drop leads to a significant reduction in the net available voltage amplitude, the optimal intermediate current trajectory sequence planning generates a more conservative optimal intermediate current trajectory sequence with a smaller upper limit of the rate of change. This ensures that the current moves towards the reference value along the fastest feasible path under the current constraints, rather than along the old, no-longer-feasible path. When the bus voltage recovers, the upper limit of the rate of change is increased accordingly to trigger replanning, generating a more aggressive optimal intermediate current trajectory sequence to fully utilize the recovered voltage margin. The amplitude-limiting smoothing process limits the deviation of the target values ​​between the old and new optimal intermediate current trajectory sequences at the switching moment, avoiding current jumps caused by sudden trajectory changes.

[0074] Because the angle index memory table is updated with the actual executed feedforward compensation voltage value, when the voltage constraint tightens, the actual executed value is scaled down and lower than the original extracted value. The updated angle index memory table stored value naturally decreases. Therefore, the historical feedforward compensation voltage vector extracted from the angle index memory table in subsequent control cycles has reflected the achievable level under the current voltage conditions, forming an adaptive closed-loop adjustment.

[0075] Therefore, under the condition of dynamic changes in voltage constraints and concurrent periodic load disturbances, deadbeat current predictive control plans the optimal intermediate current trajectory sequence and allocates voltage based on the currently available voltage space in each control cycle, tracks the reference current along the fastest feasible path allowed by the current constraints, and achieves the best dynamic response.

[0076] The following is an example of an application of the present invention, such as Figure 2-9 As shown, the implementation process is as follows: An electric vehicle is equipped with a permanent magnet synchronous motor driving an air conditioning compressor. The motor has a rated power of 3.5kW, a pole pair count of p=4, an d-axis inductance of Ld=3.2mH, a q-axis inductance of Lq=5.8mH, a stator resistance of Rs=0.18Ω, a permanent magnet flux linkage of ψf=0.085Wb, and a control period of Ts=100μs. During vehicle operation, the driver simultaneously activates the cooling mode and depresses the accelerator pedal, causing a surge in the instantaneous discharge power of the power battery, resulting in a drop in the DC bus voltage from the normal 340V to 274V. Simultaneously, the periodic load torque generated by the compressor's intake and exhaust switching persists, and the rotor operates at an electrical angular velocity ωe=628rad / s (corresponding to a speed of approximately 1500rpm). The controller executes the complete control flow of this method under this dual disturbance superposition condition.

[0077] At the current sampling moment (denoted as moment n), the controller synchronously acquires the motor operating status through the current sensor, encoder, and voltage sampling circuit. Simultaneously, it reads historical data from the sliding window and the remaining intermediate current trajectory sequence from the previous cycle from the buffer. The sliding window length Nw needs to cover at least one complete mechanical rotation cycle. At a speed of 1500 rpm, one mechanical cycle corresponds to a duration of 60 / 1500×4=0.16s (considering the number of pole pairs p=4, the mechanical cycle is 4 times the electrical cycle), corresponding to a sampling point count of Nw=0.16s / 100μs=1600 sampling points.

[0078] Table 1. Motor operating status data at the current sampling time.

[0079] The remaining intermediate current trajectory sequence cached in the previous control cycle contains three target values ​​to be executed, with the upper limit of the rate of change recorded as Δimax_prev=0.412A.

[0080] The controller calculates the current dq axis current error value based on the data in Table 1, and normalizes and maps 1600 historical error values ​​within the sliding window according to the rotor mechanical angle. It then compares the correlation between the error trajectory of the current mechanical cycle and the previous complete mechanical cycle in the angle domain.

[0081] The current q-axis current error is eq=iq iq=9.50 8.73 = 0.77A, the current d-axis current error is ed = id id= 1.00 ( 1.24)=0.24A. Differentiate the error sequence to obtain the error rate of change. The coordinates of the current phase trajectory point are (eq, rate of change of eq)=(0.77,0.031). Determine the current state point on the phase plane.

[0082] The historical error values ​​within the sliding window are rearranged according to their corresponding mechanical angles. An angle grid number of M=360 (one grid point per degree) is selected, and the trajectory point sequences of the two cycles are resampled and aligned using linear interpolation. The following are six representative angle grid point data points from the current mechanical cycle: Table 2. Alignment data of q-axis error angle between the current and previous mechanical cycles (partial).

[0083] The sequence means are: x-mean = 0.142A, y-mean = 0.136A. The normalized correlation coefficient is calculated based on all 360 grid points.

[0084] If ρ=0.969 exceeds the periodicity threshold of 0.85 and this condition has been met for 3 consecutive mechanical cycles (the number of confirmation cycles is set to 3), the controller will mark the current state as a periodic disturbance mode.

[0085] In periodic disturbance mode, the controller first calculates the predicted mechanical angle one control cycle ahead:

[0086] Using the predicted mechanical angle of 128.3° as the index, query the angle index memory table and extract the historical feedforward compensation voltage vector stored in this angle partition (corresponding to the 128°~129° partition).

[0087] Table 3. Angle Index Memory Table Lookup Results and Net Available Voltage Calculation

[0088] The current maximum available voltage amplitude is calculated from the bus voltage:

[0089] Feedforward Amplitude The net available voltage amplitude is below the maximum occupancy limit.

[0090] The controller calculates the upper limit of the rate of change of the dq-axis current based on the net available voltage amplitude Vnet = 146.1V. First, the back electromotive force voltage component is calculated:

[0091]

[0092]

[0093] Remaining available voltage amplitude:

[0094] Upper limit of current change rate:

[0095] Calculate the upper limit of the rate of change relative to the change amount:

[0096] The current bus voltage of 274V is in the low-voltage range (assuming a low-voltage range threshold of 300V). The corresponding reprogramming trigger threshold for the low-voltage range is 0.08. δ=2.153 0.08 triggers replanning of the optimal intermediate current trajectory sequence.

[0097] Based on the current measured value (id, iq) = ( Starting from 1.24, 8.73)A, the reference value is (id). iq )=( Given the terminal target (1.00, 9.50)A and the constraint Δimax = 1.299A, a cycle-by-cycle recursive calculation is performed. The current error vector magnitude between the target and the target is:

[0098] The target can be reached in one step, and the intermediate target current value can be directly taken as the reference value.

[0099] Table 4 Results of Optimal Intermediate Current Trajectory Planning

[0100] The first step target value of the old remaining trajectory sequence is ( 1.08, 9.21)A, the new initial target value is ( (1.00, 9.50)A, q-axis deviation is 9.50 9.21 = 0.29A, the smoothing limit is set to 0.35A, the deviation is not exceeded, so the new initial target value is directly adopted. The final intermediate target current value is (id_ref, iq_ref) = ( 1.00,9.50)A.

[0101] The controller uses the intermediate target current value ( Replace the original reference value with 1.00, 9.50)A, perform a two-step deadbeat prediction operation, and solve for the nominal voltage vector.

[0102] In the two-step prediction, the predicted current values ​​id(n+1) and iq(n+1) at time n+1 are jointly determined by the current measured values ​​and the voltage to be determined. Then, setting the current at time n+2 to the intermediate target value, the voltage to be applied at time n+1 is obtained through inverse solving, which is the nominal voltage vector. Substituting these values ​​into the discretized difference equation and combining them with the current parameter values, ωe = 628 rad / s is taken for both time points during the solution process. The final solution result is as follows: Table 5 Results of Two-Step Deadbeat Prediction Calculation

[0103] The nominal voltage amplitude of 118.47V is lower than the net available voltage amplitude of 146.1V, which meets the planning constraints in step 4, indicating that the intermediate target generated by the cycle-by-cycle recursive calculation is executable within the net available voltage range.

[0104] In the periodic disturbance mode, the nominal voltage vector is superimposed with the historical feedforward compensation voltage vector:

[0105]

[0106]

[0107] Comparing the synthesized voltage amplitude of 130.15V with the current maximum output voltage amplitude Umax=158.2V: 130.15V<158.2V, the synthesized voltage vector has not exceeded the limit, and the synthesized voltage vector is directly used as the final dq axis voltage command value.

[0108] Table 6. Results of Voltage Vector Synthesis and Constraint Verification

[0109] The controller will ultimately assign the dq axis voltage command value ( The voltage (9.43, 129.81)V is converted into a PWM duty cycle signal via space vector modulation and output to the inverter. Since the synthesized voltage in this control cycle does not exceed the limit, the actual feedforward compensation voltage value is the original extracted value (ud_ff, uq_ff) = ( 3.82,11.47)V.

[0110] Using the angle partition (127°~128° partition) corresponding to the current rotor mechanical angle θm=127.4° as the index, the angle index memory table is updated using a weighted average method. Let the weight coefficients be 0.2 (new value weight) and 0.8 (old value weight), and the original stored value is ( 3.75, 11.31)V:

[0111]

[0112] Table 7. Update Results of Angle Index Memory Table

[0113] At the same time, the controller will extract the remaining intermediate current trajectory sequence after the first target value (in this example, the target has been reached in one step, and the remaining sequence is empty) and the current upper limit of the rate of change Δimax=1.299A and write it into the buffer area for the next control cycle step 1 to read.

[0114] The data continuity of the entire control process is reflected as follows: Step 1, the bus voltage of 274V and rotor mechanical angle of 127.4° are the source data for all subsequent calculations; Step 2 uses the historical error sequence to calculate ρ=0.969, establishes the periodic disturbance mode marker, and determines whether to activate the feedforward extraction branch in Step 3; Step 3 extracts the feedforward voltage from the angle index memory table with the predicted angle of 128.3°, and subtracts its amplitude of 12.09V from 158.2V to obtain the net usable voltage of 146.1V, which is directly passed to Step 4; Step 4 calculates the upper limit of the rate of change of 1.299A based on 146.1V, triggering replanning to generate the intermediate target current ( 1.00, 9.50)A, input to step 5; step 5 uses this intermediate target to perform a two-step deadbeat prediction to solve for the nominal voltage ( 5.61, 118.34)V, passed to step 6; step 6 superimposes the nominal voltage and feedforward voltage and verifies, confirming that 130.15V is within the limit, and outputs the final voltage command; step 7 writes the actual executed feedforward value back to the angle index memory table according to the weighted average, and caches the upper limit of the rate of change, providing historical data for the next control cycle. The entire data link forms a complete closed loop from initial sampling to memory table update, and the values ​​between each step are strictly sequential, reflecting the coordinated control logic under the concurrent conditions of bus voltage drop and periodic load disturbance.

[0115] The embodiments of the present invention have been described above. However, these embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many equivalent embodiments based on the guidance of these embodiments, and all of these are within the protection scope of these embodiments. Specific examples have been used in this document to illustrate the principles and implementation methods of the present invention. The above examples are only for the purpose of helping to understand the method and core ideas of the present invention. The above descriptions are only preferred embodiments of the present invention. It should be noted that due to the limitations of textual expression, there are objectively infinite specific structures. For those skilled in the art, several improvements, modifications, or changes can be made without departing from the principles of the present invention, and the above technical features can also be combined in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the protection scope of the present invention.

Claims

1. A method for predicting the optimal dynamic response deadbeat current of a permanent magnet synchronous motor, characterized in that, Includes the following steps: Acquire the measured value of dq-axis current, reference value of dq-axis current, rotor mechanical angle, rotor electrical angular velocity, real-time DC bus voltage value at the current sampling moment, as well as the historical sequence of dq-axis current tracking error within the sliding window and the remaining intermediate current trajectory sequence stored in the previous control cycle; Based on the historical sequence of the dq axis current tracking error, the error value is normalized and mapped according to the rotor mechanical angle. The normalized correlation coefficient between the trajectory point sequence of the current mechanical cycle and the corresponding angle position of the previous complete mechanical cycle is calculated. According to the comparison result of the normalized correlation coefficient and the periodicity determination threshold, the current state is marked as a periodic disturbance mode or a non-periodic disturbance mode. In periodic disturbance mode, the historical feedforward compensation voltage vector is extracted from the angle index memory table based on the predicted mechanical angle, and its magnitude is subtracted from the current maximum available voltage vector magnitude to obtain the net available voltage magnitude; in non-periodic disturbance mode, the current maximum available voltage vector magnitude is directly used as the net available voltage magnitude. The upper limit of the dq axis current change rate is calculated based on the net available voltage amplitude. Based on the comparison between the relative change of the upper limit of the change rate and the replanning trigger threshold, it is decided to extract the intermediate target current value from the remaining intermediate current trajectory sequence or to perform a cycle-by-cycle recursive calculation to generate a new optimal intermediate current trajectory sequence and extract the intermediate target current value. Using the intermediate target current value as the objective, a two-step deadbeat prediction operation is performed based on the discretized current prediction model of permanent magnet synchronous motor to calculate the nominal voltage vector; In periodic disturbance mode, the nominal voltage vector is superimposed with the historical feedforward compensation voltage vector to generate a composite voltage vector. After performing amplitude constraint verification on the composite voltage vector, the final dq-axis voltage command value is generated. In non-periodic disturbance mode, the nominal voltage vector is performed to generate the final dq-axis voltage command value. The final dq axis voltage command value is output to the inverter after space vector modulation, and the angle index memory table is updated with the actual feedforward compensation voltage value in periodic disturbance mode.

2. The method for predicting the best dynamic response deadbeat current of a permanent magnet synchronous motor according to claim 1, characterized in that, The normalized correlation coefficient is calculated using the Pearson correlation coefficient formula, specifically as follows: the error values ​​at each angle grid point in the error sequence after angle alignment within the current machine cycle and the error sequence at the corresponding angle position within the previous complete machine cycle are subtracted from the mean of their respective sequences, and then multiplied point by point and summed. The sum is then divided by the product of the square roots of the sum of the squares of the mean deviations of the two sequences, and the quotient is the normalized correlation coefficient. When the normalized correlation coefficient exceeds the periodicity determination threshold and the number of consecutive mechanical cycles that meet this condition reaches the number of confirmed cycles, the current state is marked as a periodic disturbance mode.

3. The method for predicting the best dynamic response deadbeat current of a permanent magnet synchronous motor according to claim 1, characterized in that, The normalization mapping refers to rearranging the error values ​​at each sampling time from the time domain to the mechanical angle domain using the corresponding rotor mechanical angle as an index; When the rotational speed changes, resulting in different numbers of sampling points within the same mechanical angle range, a linear interpolation algorithm is used to resample the trajectory point sequences of the two cycles in the angle domain, so that the trajectory point sequences are aligned on the same angle grid.

4. The method for predicting the best dynamic response deadbeat current of a permanent magnet synchronous motor according to claim 1, characterized in that, The predicted mechanical angle is the current rotor mechanical angle plus the rotor electric angular velocity divided by the number of motor pole pairs and then multiplied by the control cycle duration to obtain the angle increment. The angle index memory table uses the rotor mechanical angle as the index to uniformly discretize a complete mechanical rotation cycle into multiple angle partitions. Each angle partition stores the d-axis and q-axis components of the feedforward compensation voltage that was most recently executed in that partition. When the amplitude of the historical feedforward compensation voltage vector exceeds the product of the current maximum available voltage vector amplitude and the upper limit of the occupancy ratio, the historical feedforward compensation voltage vector is scaled proportionally according to the upper limit of the occupancy ratio, and the scaled amplitude is deducted from the current maximum available voltage vector amplitude as the feedforward occupancy amount.

5. The method for predicting the best dynamic response deadbeat current of a permanent magnet synchronous motor according to claim 1, characterized in that, The calculation process for the upper limit of the dq axis current change rate is as follows: based on the net available voltage amplitude, the amplitude of the back electromotive force voltage component determined by the rotor electric angular velocity, dq axis inductance, the current measured value of dq axis current and the permanent magnet flux linkage is subtracted to obtain the remaining available voltage amplitude. The upper limit of the dq-axis current change rate is obtained by dividing the product of the remaining available voltage amplitude and the control cycle duration by the square root of the sum of the squares of the d-axis inductance and the q-axis inductance. The relative change of the upper limit of the change rate is the ratio of the absolute value of the difference between the current upper limit of the change rate and the upper limit of the change rate of the previous control cycle to the upper limit of the change rate of the previous control cycle.

6. The method for predicting the best dynamic response deadbeat current of a permanent magnet synchronous motor according to claim 1, characterized in that, When the relative change of the upper limit of the rate of change exceeds the replanning trigger threshold, starting from the current measured value of the dq-axis current, taking the updated upper limit of the rate of change of the dq-axis current as the constraint and the reference value of the dq-axis current as the terminal target, a new optimal intermediate current trajectory sequence is generated by recursive calculation on a cycle-by-cycle basis. The first step target value is extracted from the new optimal intermediate current trajectory sequence. The deviation between the first step target value and the target value at the corresponding time in the old remaining intermediate current trajectory sequence is subjected to amplitude limiting and smoothing to obtain the smoothed intermediate target current value. The aforementioned amplitude limiting and smoothing process refers to truncating the deviation to the smoothing amplitude limit when the absolute value of the deviation exceeds the smoothing amplitude limit, and using the old target value plus the truncated deviation value as the intermediate target current value. When the absolute value of the deviation does not exceed the smoothing amplitude limit, the initial target value is directly used as the intermediate target current value. When the remaining intermediate current trajectory sequence is empty and the relative change of the upper limit of the rate of change does not exceed the replanning trigger threshold, the cycle-by-cycle recursive calculation is also performed to generate a new optimal intermediate current trajectory sequence and the first target value is extracted as the intermediate target current value.

7. The method for predicting the best dynamic response deadbeat current of a permanent magnet synchronous motor according to claim 1, characterized in that, The process of the two-step deadbeat prediction operation is as follows: Based on the forward Euler discretization difference equation of the dq-axis voltage equation of the permanent magnet synchronous motor, the dq-axis current value after the first step is predicted using the current measured dq-axis current value as the initial value and the nominal voltage vector to be determined. Then, the dq-axis current value after the first step is predicted using the predicted dq-axis current value as the initial value. The predicted dq-axis current value after the second step is set to be equal to the intermediate target current value. The d-axis and q-axis components of the nominal voltage vector to be applied are then calculated in reverse.

8. The method for predicting the best dynamic response deadbeat current of a permanent magnet synchronous motor according to claim 1, characterized in that, The amplitude constraint verification process is as follows: When the magnitude of the synthesized voltage vector exceeds the current maximum output voltage vector magnitude, the scaling factor is calculated as the ratio of the current maximum output voltage vector magnitude to the synthesized voltage vector magnitude. The nominal voltage vector and the historical feedforward compensation voltage vector are scaled proportionally by the scaling factor and then added together to generate the final dq axis voltage command value. When the scaling factor is lower than the minimum scaling threshold, the amplitude of the historical feedforward compensation voltage vector is reduced by a preset reduction ratio and then superimposed with the nominal voltage vector to generate the final dq axis voltage command value. The reduced historical feedforward compensation voltage vector is used as the feedforward value actually used in the current control cycle for updating the angle index memory table.

9. The method for predicting the best dynamic response deadbeat current of a permanent magnet synchronous motor according to claim 1, characterized in that, The angle index memory table is updated using a weighted average method, which involves weighting and fusing the feedforward compensation voltage value actually executed in the current control cycle with the original stored value in the angle index memory table according to a fixed weighting coefficient, and then writing it back to the corresponding angle partition. When the current state switches from aperiodic disturbance mode to periodic disturbance mode, or when the change in DC bus voltage between adjacent control cycles exceeds the voltage jump threshold, the values ​​of all storage units in the angle index memory table are attenuated according to a preset attenuation coefficient.

10. A system for predicting the best dynamic response deadbeat current for a permanent magnet synchronous motor, used to execute the method for predicting the best dynamic response deadbeat current for a permanent magnet synchronous motor as described in any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to acquire the measured value of the dq axis current, the reference value of the dq axis current, the rotor mechanical angle, the rotor electrical angular velocity, the real-time DC bus voltage value, the historical sequence of the dq axis current tracking error within the sliding window, and the remaining intermediate current trajectory sequence stored in the previous control cycle. The periodic disturbance detection module is used to normalize the error value according to the rotor mechanical angle based on the historical sequence of the dq axis current tracking error, calculate the normalized correlation coefficient between the trajectory point sequence of the corresponding angle position of the current mechanical cycle and the previous complete mechanical cycle, and mark the current state as a periodic disturbance mode or a non-periodic disturbance mode according to the comparison result of the normalized correlation coefficient and the periodicity determination threshold. The net available voltage calculation module is used to extract the historical feedforward compensation voltage vector from the angle index memory table based on the predicted mechanical angle under periodic disturbance mode and subtract its magnitude from the current maximum available voltage vector magnitude to obtain the net available voltage magnitude. Under non-periodic disturbance mode, the current maximum available voltage vector magnitude is directly used as the net available voltage magnitude. The trajectory planning module is used to calculate the upper limit of the dq axis current change rate based on the net available voltage amplitude, and extract the intermediate target current value from the remaining intermediate current trajectory sequence or perform cycle-by-cycle recursive calculation to generate a new optimal intermediate current trajectory sequence and extract the intermediate target current value based on the comparison result of the relative change of the upper limit of the change rate and the replanning trigger threshold. The deadbeat prediction module is used to perform two-step deadbeat prediction operations based on the discretized current prediction model of permanent magnet synchronous motor with the intermediate target current value as the target, and calculate the nominal voltage vector. The voltage synthesis and constraint verification module is used to superimpose the nominal voltage vector and the historical feedforward compensation voltage vector to generate a synthesized voltage vector under periodic disturbance mode, and then perform amplitude constraint verification to generate the final dq axis voltage command value. Under non-periodic disturbance mode, the module performs amplitude constraint verification on the nominal voltage vector to generate the final dq axis voltage command value. The output and update module is used to output the final dq axis voltage command value to the inverter after space vector modulation, and update the angle index memory table with the actual feedforward compensation voltage value executed in the periodic disturbance mode.

Images