A method and system for controlling current of a surface-mounted permanent magnet synchronous motor
By calculating the prefix voltage action of executed switching events using the same state transition exponent, a robust reachable domain and PWM edge are formed in the surface-mounted permanent magnet synchronous motor, solving the delay and loss problems of current control and achieving stable current control.
Patent Information
- Application Number
- CN202611505123.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-24
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies for controlling the current of surface-mounted permanent magnet synchronous motors suffer from calculation delays and switching losses, and cannot effectively ensure that switching events are calculated in unison with the actual voltage, leading to unstable control modes.
The prefix voltage action of the executed switching event is calculated using the same state transition exponent. A robust reachable domain is formed by the remaining time and the measured bus voltage. PWM edges are generated in the exponential time coordinate. The actual action is verified after timer quantization to ensure the accuracy of the closure criterion.
It effectively reduces computational latency, lowers switching losses, and ensures the stability and accuracy of current control, making it suitable for implementation in DSPs, FPGAs, or programmable timers.
Smart Images

Figure CN123122340A_ABST
Abstract
Description
Technical Field
[0001] This invention proposes a method and system for controlling the current of a surface-mounted permanent magnet synchronous motor, relating to the fields of digital current control of permanent magnet synchronous motors, space vector pulse width modulation, and electronic speed controllers for unmanned aerial vehicles. Background Technology
[0002] The current loop of a permanent magnet synchronous motor is the foundation for torque control, speed control, and position control. Digital controllers typically sample the stator current at carrier peaks, valleys, or fixed sampling points, and update the PWM comparator register after coordinate transformation, current regulation, and pulse width modulation calculations. The wait between sampling, calculation, and register activation introduces a delay in new voltage commands relative to the sampled current; increasing the switching frequency of power devices can reduce the proportion of delay, but it also increases switching losses and thermal design stress.
[0003] Existing intra-carrier instantaneous update methods typically utilize the zero-voltage vector duration to accommodate control calculations and update the PWM immediately after calculation. When the zero-voltage vector time is insufficient, a switch to another sampling or update timing can be initiated. This type of method can reduce the equivalent delay in some operating conditions, but the switching voltages already executed during the calculation are usually still treated as timing conditions, and their actual effect on the current at the end of the current sub-cycle is not uniformly calculated with subsequent voltages. Under high-modulation conditions, bus dips, or fluctuations in calculation time, the zero-vector window shortens, and the control mode or timing conditions are prone to change.
[0004] Predictive current control and deadbeat current control can calculate the reference voltage, candidate voltage vector, and their duration based on the motor model, and calculate the delay through one or more steps of prediction compensation. However, if only the current at the next sampling moment is predicted, only the duration of the candidate vector is optimized, or only the voltage of the previous cycle is substituted into the delay compensation model, it cannot be proven that all the switching events that have actually occurred between sampling and calculation completion in the current sub-cycle and the new edge after calculation are closed on the same physical integral.
[0005] Furthermore, the continuously calculated vector action time must ultimately be quantized into a timer clock count value and is subject to minimum pulse width, dead time, and register load time constraints. Near the boundary of the remaining reachable domain, continuously feasible actions may cross the robust boundary after rounding. If feasibility is only determined within the continuous domain without recalculating the action based on the quantized actual switching sequence, it cannot be guaranteed that the edge actually delivered to the power device still satisfies the current endpoint and voltage reachability constraints.
[0006] Therefore, a current control method is needed that uses the same state transition exponent kernel to run through target calculation, executed prefix calculation, remaining reachable projection, edge generation, quantization verification, and endpoint feedback. This method would transform the physical effect of computation time from unusable pure waiting to an allocation problem between executed action and remaining reachable capability, and make the actual loaded timer edge, rather than the continuous theoretical duty cycle, the final object of the closure criterion. Summary of the Invention
[0007] In view of this, in order to fill the gaps and deficiencies in the prior art, the present invention proposes a method and system for controlling the current of a surface-mounted permanent magnet synchronous motor, in order to solve the problems that have occurred in the background art.
[0008] This invention proposes a method and system for controlling the current of a surface-mounted permanent magnet synchronous motor, comprising the following:
[0009] According to a first aspect of the present invention, a method for controlling the current of a surface-mounted permanent magnet synchronous motor is provided, characterized by comprising the following:
[0010] In the The start time of each current control sub-cycle The stator current of the surface-mount permanent magnet synchronous motor is sampled and recorded by the switching event recording unit from... Until the calculation is completed The actual inverter switching states that have been executed, the DC bus voltage corresponding to each switching state, and the switching edge times;
[0011] The state transition index kernel is determined based on the stator resistance and stator inductance. Using the same state transition index kernel, the target total voltage action required for the current current control sub-cycle is calculated from the target endpoint current. Simultaneously, the executed prefix voltage action is calculated from the already executed inverter switching states, DC bus voltage, and switching edge moments. The result is obtained by subtracting the executed prefix voltage action from the target total voltage action. To the end of the sub-cycle The remaining action gap;
[0012] according to to The remaining nominal reachable action domain is constructed by the convex hull of the remaining exponential time measure, the current measured DC bus voltage, and the inverter basic voltage vector. The remaining nominal reachable action domain is then eroded by the uncertain action formed by motor parameter error, back EMF action estimation error, bus voltage error, dead zone voltage drop, and sampling error to obtain the remaining robust reachable action domain.
[0013] The remaining action gap is projected onto a continuous executable domain located inside the remaining robust reachable action domain to obtain the remaining action to be executed. Within the exponential time coordinate determined by the state transition exponential kernel, the remaining action to be executed is decomposed into exponential time weights of adjacent active voltage vectors and zero voltage vectors, and generated through the inverse transformation of the exponential time coordinate. The PWM edge timing that has not yet occurred;
[0014] The PWM edge time is quantized into a timer clock count value. Based on the quantized actual switching state sequence and the quantized edge time, the quantized actual residual action is recalculated using the same state transition exponent kernel. The timer clock count value is allowed to be loaded only if the quantized actual residual action is within the residual robust reachable action domain and its closure residual relative to the remaining action to be executed is not greater than the preset action residual upper limit. Otherwise, the combination with the smallest closure residual within the residual reachable action domain is selected from adjacent timer clock count combinations.
[0015] Execute the selected PWM edge during the remaining time of the current current control sub-cycle, and... The stator current at the sampling endpoint is sampled, and a physical closure residual is generated based on the difference between the stator current at the sampling endpoint and the predicted endpoint current corresponding to the actual remaining action after quantization. The physical closure residual is then used to update the uncertain action of the next current control sub-cycle or the shrinkage margin of the continuous executable domain.
[0016] Furthermore, the surface-mounted permanent magnet synchronous motor in The current model in the stationary coordinate system, the precise state transition relation, and the state transition exponent kernel are respectively
[0017] ; (1);
[0018] ; (2);
[0019] (3);
[0020] ;(4);
[0021] in, for Shaft equivalent stator inductance, in H; Stator phase resistance, unit: ;
[0022] , and These are the stator current vector, inverter output voltage vector, and back EMF vector, respectively, with units of A, V, and V;
[0023] The natural decay rate of current is expressed in s⁻¹. The natural decay coefficient of the current in a current-controlled sub-cycle;
[0024] It is a dimensionless state transition exponent kernel; and These are the voltage action and the back electromotive force action, respectively, both in V·s.
[0025] Furthermore, regarding from to The Each actual switch state range Define an exponential time measure and calculate the effect of the executed prefix voltage:
[0026] ; (5);
[0027] ;(6);
[0028] in, This represents the number of actual switch state intervals within the prefix. For the first One actual switch state;
[0029] The actual voltage vector, in units of V, is obtained by combining the measured DC bus voltage, bridge arm switch status, and dead zone current direction in this section.
[0030] The unit is s; The unit is V·s.
[0031] Furthermore, the target endpoint current is At that time, the target total voltage action, the residual action gap, and the residual exponential time measure are respectively
[0032] ;(7);
[0033] ;(8);
[0034] The remaining nominally reachable scope, the remaining robustly reachable scope, and the contiguous executable scope are respectively:
[0035] (9);
[0036] in, This is an estimated value of the back electromotive force, in V·s; The target total voltage effect, in V·s; The remaining action gap is expressed in V·s;
[0037] Remaining physical time, in seconds; to These are the eight basic voltage vectors of a three-phase two-level inverter, in units of V;
[0038] This is the set of uncertain action quantities calculated from parameters, back electromotive force, bus, dead zone, and sampling error, in V·s;
[0039] A set of variables reserved for timer quantization and minimum pulse width, in V·s;
[0040] This is the Minkowski set difference operation.
[0041] Furthermore, the remaining action to be executed is determined by the following formula:
[0042] ;(10);
[0043] in, This is the positive definite current error weight matrix; when using a contraction line pointing from the current current to the target endpoint current, it is defined as follows:
[0044] (11);
[0045] And will make The largest within the contiguous executable domain Determined as the maximum feasible shrinkage factor ;
[0046] When the continuous executable domain adopts a hexagonal half-space Indicates and When located within the contiguous executable domain, the following is obtained:
[0047] (12);
[0048] in, This is the upper limit of the instruction shrinkage factor; For hexagonal number 1 The unit outward normal vector of each face; The shrinkage boundary of the face corresponding to the continuous executable domain, in units of V·s; This is the interval truncation operator.
[0049] Furthermore, the exponential time coordinate, the decomposition of the remaining action to be executed, and the inverse edge transform are respectively:
[0050] ;(13);
[0051] ;(14);
[0052] (15);
[0053] in, The sector number of the voltage vector containing the remaining action to be executed; and These are the two adjacent active voltage vectors of this sector; , and These are the exponential time weights of the zero voltage vector and two adjacent active voltage vectors, respectively, in seconds; A dimensionless exponential time coordinate; The edge time to be quantized is in seconds (s).
[0054] Furthermore, let the timer clock period be... The quantized edge count, quantized edge time, and quantized residual action are respectively...
[0055] (16)
[0056] ;(17);
[0057] The conditions for implementing the license are as follows
[0058] (18)
[0059] in, A quantization operator for selecting timer count values within the continuous executable domain; This represents the number of actual switching state intervals after quantization. The actual residual action after quantification, in V·s;
[0060] To quantify the closed residual of the executed action relative to the continuous actions to be executed, the unit is V·s; The upper limit of the allowable quantization action error, in V·s;
[0061] This is the minimum effective pulse width allowed for power devices and drive circuits, measured in seconds (s).
[0062] According to a second aspect of the present invention, the present invention provides a system for controlling the current of a surface-mount permanent magnet synchronous motor, for executing a method for controlling the current of a surface-mount permanent magnet synchronous motor as described in any one of the present invention, characterized in that it includes a three-phase two-level inverter, a current sampling circuit, a DC bus voltage sampling circuit, a processor, a memory, and a PWM timer, a switching event recording unit, and an intra-carrier edge scheduling unit; wherein the switching event recording unit is used to record the inverter switching states, corresponding DC bus voltages, and edge times that have been actually executed from the start of the current control sub-cycle to the completion time of the calculation; the intra-carrier edge scheduling unit is used to load the remaining PWM edges that have been quantized and verified.
[0063] Furthermore, the surface-mounted permanent magnet synchronous motor is a surface-mounted permanent magnet synchronous motor driven by a three-phase two-level inverter powered by a DC bus.
[0064] Furthermore, it also includes an unmanned aerial vehicle (UAV) comprising a fuselage, a propeller, a surface-mounted permanent magnet synchronous motor, and an electronic speed controller, wherein the electronic speed controller utilizes the executed prefix voltage action and the remaining robust reachable action domain of the current current control subcycle to generate and execute PWM edges that have not yet occurred in the current current control subcycle without switching current control modes.
[0065] The present invention has the following advantages:
[0066] (1) The executed prefix is no longer considered an unusable computational dead zone, but instead enters the current endpoint state equation according to the actual switching events and measured bus voltage, avoiding the substitution of the real prefix with the nominal one-beat voltage.
[0067] (2) The target action, prefix action, back potential action, remaining reachable domain, quantized action and endpoint residual all use the same state transition exponent kernel to avoid inconsistencies between the accurate model and ordinary volt-second modulation.
[0068] (3) The size of the reachable domain is changed by measuring the remaining exponential time at the time of completion of the calculation. There is no need to compare the calculation time with the zero vector threshold, nor is it necessary to switch between the two control modes of instant update and dual sampling and dual update.
[0069] (4) The continuous action is directly transformed into the physical edge within the carrier through the exponential time coordinate, and the actual action is recalculated after timer quantization, so that the closure criterion falls on the actual loaded count value and switching sequence.
[0070] (5) Uncertainty erosion, quantization margin, minimum pulse width and closure residual all act on the same residual domain, which can maintain the feasibility criterion in the case of bus drop, parameter perturbation and edge rounding.
[0071] (6) The required online calculations mainly include prefix interval summation, hexagonal surface constraint, two-dimensional projection or one-dimensional ray intersection, three-vector decomposition and finite neighborhood counting search, which are suitable for implementation by DSP, FPGA or programmable timer with switching event capture and intra-carrier comparison event update capabilities. Attached Figure Description
[0072] Figure 1 This is a schematic diagram of the overall closed-loop action quantity within the same exponential nuclear carrier of the present invention.
[0073] Figure 2 This is a schematic diagram of the physical mapping of the total action, executed prefix and remaining action of the target under the same exponential kernel of the present invention.
[0074] Figure 3 This is a schematic diagram of the exponential time action decomposition, timer quantization, and post-quantization verification process of the present invention. Detailed Implementation
[0075] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.
[0076] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0077] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention; as used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise; furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.
[0078] like Figures 1 to 3 As shown, this invention proposes a method and system for controlling the current of a surface-mounted permanent magnet synchronous motor, comprising the following:
[0079] The purpose of this invention is to solve the following interrelated technical problems: how to treat the inverter voltage actually applied during the calculation as part of the target action of the current sub-cycle, instead of discarding it as an invalid delay; how to determine whether the target endpoint current is still achievable in the current sub-cycle based on the remaining time and the current bus voltage; and how to ensure that the final loaded PWM edge is still within the physically reachable boundary when timer quantization, minimum pulse width, and uncertainty exist.
[0080] This invention does not select the control mode based on the relationship between the calculation time and the zero vector duration. Instead, it starts from the motor physical model in equation (1) and uses the same exponent in equation (3) to calculate the entire control sub-cycle. The target endpoint current corresponds to a target total voltage action; the actual switching history from the sampling time to the calculation completion time corresponds to the executed prefix voltage action; the difference between the two is the action gap that actually needs to be filled in the remaining time.
[0081] The residual action is not an arbitrary two-dimensional vector, but must lie within the residual reachable domain formed by the residual exponential time measure and the inverter voltage hexagon. Through uncertain action erosion and quantization margin shrinkage, a continuous executable domain is first obtained, and then the action gap is projected onto this domain. The projection result is assigned action weights to adjacent active vectors and zero vectors within the exponential time coordinate, and the true physical edge is obtained through inverse logarithmic mapping.
[0082] After consecutive edges are converted into timer counts, the invention again calculates the action based on the quantized actual switching sequence. Execution is only permitted if the quantized action simultaneously satisfies the robust reachability constraint, minimum pulse width constraint, and closed residual constraint; otherwise, a feasible solution is selected from adjacent clock combinations. The next sampled current forms the endpoint physical residual and is used to update the uncertain action of the next sub-cycle. Thus, the target, prefix, reachability region, edge, register, and next measurement belong to the same causal closed loop.
[0083] Unless otherwise specified, seconds are used consistently when entering time into the formula; microsecond values are multiplied by before calculation. Voltage action quantities are uniformly expressed in V·s; Vectors employ magnitude-invariant coordinate transformations; set boundaries, projections, and quantization verifications use the same coordinate transformation scale.
[0084] The relevant appendices are as follows:
[0085] Table 1. Main Symbols and Physical Meanings
[0086]
[0087] Table 2. Main Symbols and Physical Meanings
[0088]
[0089] Table 3. Main Symbols and Physical Meanings
[0090]
[0091] The technical solution adopted in this invention is as follows: A current control method for a surface-mounted permanent magnet synchronous motor based on the same exponent kernel prefix action calculation and intra-carrier executable closure, comprising the following steps:
[0092] Step 1: Core Original Physical Closure Relationship:
[0093] Surface-mounted permanent magnet synchronous motor An inductor in a stationary coordinate system can be considered isotropic, and its current model is given by equation (1). By precise integration, we obtain equations (2) to (4). The physical quantity that determines the endpoint current is the integral of the voltage and back EMF under the same exponential kernel over the entire sub-cycle, rather than the calculation completion time tag of a certain voltage command.
[0094] The switching events that have already occurred have definite edge times and switching states. Each actual switching state interval can be converted into voltage action quantities and summed using equations (5) and (6). If the actual bus voltage changes within the prefix, the bus voltage latched or interpolated within that interval is used for each switching interval; if the output voltage in the dead zone is inconsistent with the logic state of the bridge arm, the actual voltage vector is reconstructed according to the phase current direction, and the unmodeled part enters the set of uncertain action quantities.
[0095] The target endpoint current is converted into the target total voltage effect using equation (7). In equation (7)... Since the voltage action uses the same exponent kernel, the back EMF is not treated as an external empirical feedforward, but rather as part of the state transition identity. The remaining action gap is the target total action minus the executed prefix action.
[0096] Step 2: Remaining robust reachable region and endpoint projection:
[0097] from arrive The entire achievable continuous average voltage belongs to the convex hull of the basic voltage vector of the two-level inverter. Multiply this convex hull by the residual exponential time measure. Thus, the remaining nominal reachable domain of action of equation (9) is obtained. Using constant amplitude... During the transformation, the reachable region is a hexagon centered at the origin, and its half-space form is:
[0098] ;(19);
[0099] in, This is the current measured DC bus voltage, in volts (V). For the remaining nominally reachable domains The boundary of each face, in units of V·s.
[0100] Resistance, inductance, back EMF, bus, dead zone, and sampling error are uniformly converted into a set of action quantities. The remaining nominally reachable region is obtained by performing Minkowski erosion on the remaining region. Then, jitter is generated using a timer clock, minimum pulse width, and edge loading. The continuous executable domain is obtained. The projection of Equation (10) directly targets the endpoint current error; when the target remaining action is already within the continuous executable domain, the projection remains unchanged; when it is outside the domain, the projection returns the executable action that is closest to the target endpoint current in the current remaining time.
[0101] Equations (11) and (12) provide a one-dimensional shrinkage implementation with lower computational cost. This implementation intersects the hexagon with the action ray pointing from the current current to the target endpoint current, requiring only the comparison of the six face normal inner products and scalars. If the ray origin is not in the continuous executable domain due to prefix action anomalies, the two-dimensional projection of Equation (10) is used directly without switching to another current controller.
[0102] Ordinary SVPWM performs volt-second balance within the physical time coordinate; the precise state transition relationship of equation (2) requires the execution of an exponentially weighted action. Equation (13) maps physical time to an exponential time coordinate, within which... Equal to uniform increment Therefore, the action weights obtained from equation (14) can be directly inversely transformed into the true edge time using equation (15).
[0103] Step 3: Generation of exponential time edges and post-quantization certificate:
[0104] Ordinary SVPWM performs volt-second balance within the physical time coordinate; the precise state transition relationship of equation (2) requires the execution of an exponentially weighted action. Equation (13) maps physical time to an exponential time coordinate, within which... Equal to uniform increment Therefore, the action weights obtained from equation (14) can be directly inversely transformed into the true edge time using equation (15).
[0105] Equation (16) is not a regular undirected nearest integer. First, generate the nearest count value, and then calculate the actual switching sequence action corresponding to the count value according to equation (17). If equation (18) is not satisfied, enumerate a finite number of clock count combinations adjacent to each internal edge, eliminate combinations with unacceptable minimum pulse width and robust out-of-domain combinations, and select the combination with the smallest action closure residual from the remaining combinations. The output of this step is a timer count value that can be directly loaded, rather than a continuous duty cycle that still needs secondary interpretation.
[0106] when As it approaches zero, the exponential time measure continuously degenerates into ordinary time:
[0107] (20)
[0108] Therefore, this method naturally approaches ordinary SVPWM when the electrical attenuation is very small, without producing structural switching.
[0109] Step 4: Closure residuals, endpoint error, and robust limits:
[0110] When the actual effect after quantification satisfies equation (17), let , and Let represent the prefix action, residual action, and back EMF action formed by the actual bridge arm voltage and the actual back EMF, respectively. Define the total action residual, which includes errors in prefix reconstruction, quantization execution, and back EMF estimation.
[0111] ;(twenty one);
[0112] in, The predicted endpoint current corresponding to the action to be performed, in amperes (A); This is the upper bound of the total action residual, which includes model error, quantization error, and execution error, in V·s.
[0113] The minimum stator inductance within the allowable parameter range, expressed in ohms (H).
[0114] For a fixed current reference and using the contraction endpoint of equation (11), define ,but
[0115] ;(twenty two);
[0116] in, This is the positive lower bound of the maximum feasible contractility factor; This is a unified upper bound for the total action residual. Equations (21) and (22) show that the feasible contraction amount is mainly affected by the remaining reachable domain at the time of calculation completion, while the endpoint closure accuracy is directly determined by the actual action residual.
[0117] When the electric angular velocity is approximately constant within one control sub-cycle, the complex vector of the back electromotive force and its analytical action are:
[0118] ; (twenty three);
[0119] in, The imaginary unit; Electric angular velocity, in rad / s; This refers to permanent magnet flux linkage, measured in Wb. for Electrical angle at a given moment, in rad; for The complex representation of back electromotive force, in units of V; This is a complex representation of the action of the back electromotive force, with units of V·s. When the velocity changes significantly, the observed back electromotive force samples can be integrated piecewise using the same exponent without altering the prefix calculation and the residual closed structure.
[0120] In addition to the above, the present invention also has related embodiments, including the following:
[0121] In one embodiment of the present invention, the in-carrier switching event calculation and execution sequence includes the following:
[0122] Center-aligned PWM peak or valley triggers current sampling and latching simultaneously. Current bridge arm switch status and DC bus voltage. The switch event recording unit starts from... Start recording the changed bridge arm states and timer counts; if the switch state does not change, the prefix contains only one interval.
[0123] The processor calculates the target total action of equation (7) based on the endpoint current reference and the back electromotive force of equation (23). At the end of the calculation, the data is read... The event table is used to calculate the exponential time measure interval by interval according to equation (5), and the executed prefix action is obtained according to equation (6). If the phase current direction within the dead zone is uncertain, the upper bound of the corresponding dead zone voltage error is added. .
[0124] according to , and current Form the remaining hexagon; according to the current temperature range Magnetic saturation within the allowable range Back potential observation residuals and ADC error formation ; formed according to timer clock, edge-loaded jitter and minimum pulse width First execute equation (10) or equation (12), then execute equations (14) through (18). The selected count value is loaded through the shadow register or the intra-carrier comparison event unit, and past edges are not written back.
[0125] exist The current is sampled again. Calculated from the quantized event table. The predicted endpoint current is obtained according to equation (2), and the residual is formed by subtracting the predicted endpoint current from the measured endpoint current. The residual can be updated using the sliding maximum value, confidence upper bound, or interval envelope. When residual difference increases abnormally, increase the shrinkage margin or decrease it. However, it does not change the closed structure of the same exponential core.
[0126] In one embodiment of the present invention, the in-carrier current endpoint response mechanism level simulation includes the following:
[0127] Pick , , , back electromotive force equivalent value The upper limit of the endpoint contraction factor is taken as The current reference jumps from 0 A to 18 A in the 6th sub-cycle, and the calculation completion time is taken as 2 seconds. 7 and 12 .
[0128] By simulating the direct integral of equation (1), the exponent kernel of equation (3) is used for both the prefix and the remaining interval. Under the three computation completion times, the feasible region response of this invention coincides, 12 The peak and final values are both 18.000 A. When prefix calculation is cancelled but a new action is still applied within the remaining interval, 12 The peak value under operating conditions is 18.874 A, while the simulation endpoint remains at 18.301 A. This result verifies the necessity of the prefix action in the endpoint state identity, but does not represent a guarantee of bandwidth or torque performance in the actual machine.
[0129] In one embodiment of the present invention, the remaining reachable projection and post-quantization verification include the following:
[0130] Pick , , , , And a margin of 3.5% of the remaining hexagonal surface boundary is reserved for uncertain action. The magnitude of the representative remaining action gap is 0.4861 mV·s, which is outside the remaining robust reachable domain; after projection by equation (10), the magnitude of the action to be executed is 0.4382 mV·s.
[0131] When the 10 ns timer clock is used to directly round the edge of equation (15), the quantization action crosses the robust domain boundary. After using the inward finite neighborhood search of equations (16) to (18), the quantization action re-enters the robust domain, and the error of the quantization action relative to the continuous projected action is 0.2614. In the simulation, all geometric quantities use the same constant amplitude as in equation (1). Coordinate scale.
[0132] In one embodiment of the present invention, exponential time-allocated ablation includes the following:
[0133] Pick Voltage vector amplitude 24 V The target active vector action is taken as 62% of the exponential time measure of the entire sub-cycle. Under the same target action, the ordinary time allocation that takes 62% directly as the proportion of physical time is compared with the exponential time allocation obtained from equations (13) to (15).
[0134] when At that time, the error in the endpoint current action caused by normal time allocation is 53.58 mA; when At that point, the error increased to 123.67 mA. The numerical error of the exponential time allocation was less than... mA. Simulations show that if the target action uses an exact state transition kernel while the edges are still allocated according to ordinary physical time volt-seconds, a systematic residual will form between the theoretical model and the actual execution.
[0135] In one embodiment of the present invention, the calculation of the completion time and the feasible contraction factor includes the following:
[0136] Pick , , , The calculation completion time is 0.5 seconds. Scan to 18 A representative two-dimensional current, back EMF, and actual prefix average voltage are used, with a robust margin of 2.5% reserved at the hexagonal surface boundary.
[0137] At each delay point, the present invention first calculates based on the remaining domain. Then, using the corresponding feasible endpoint as the error benchmark, the error of the maximum endpoint identity after prefix calculation is less than... mA; with prefix accounting disabled, the maximum endpoint error is 340.52 mA. As the calculation completion time approaches the end of the sub-cycle, the remaining reachable domain shrinks. In 18 The time decreased to 0.0133. This result indicates that the computation time did not disappear, but rather the remaining reachability limited the amount of contraction achievable in this beat.
[0138] In one embodiment of the present invention, the repeated random verification of the timer quantization boundary includes the following:
[0139] A fixed random seed 20260821 was used to perform 1200 sets of mechanism-level repeated computations. The parameter range for each set is as follows: ~54 V, ~0.32 , ~0.62 mH, ~36 , for The timer clock is 5 ns, 10 ns, or 20 ns, and the continuous target action is concentrated in the range of 0.005% to 0.3% inside the robust domain boundary to verify the most dangerous boundary conditions.
[0140] The robust domain out-of-bounds rate with nearest-clock rounding is 2.58%; after adding inward neighborhood search and verification with equation (18), the out-of-bounds rate is 0. The root mean square errors of the quantization errors of the two are respectively:
[0141] 0.1666 and 0.1717 ,
[0142] This demonstrates that the present invention obtained a boundary executability certificate at the cost of very small quantization error.
[0143] The analysis examines the results of canceling prefix accounting, canceling exponential time allocation, and canceling post-quantization verification. These three ablation methods correspond to different physical stages and cannot be substituted for one another: prefix accounting ensures the total target action is consistent with the already occurred switching events; exponential time allocation ensures the continuous action is consistent with the true edge; and post-quantization verification ensures the continuous edge is consistent with the actual timer count.
[0144] The simulations described above are all mechanism-level numerical verifications based on equations (1) to (23), used to prove the consistency between the formula chain, unit relations, reachable domain, and quantization execution, and are not actual machine test results. In actual engineering implementation, the specific power device, timer resolution, motor parameter range, back EMF observation error, and sampling link should be determined accordingly. , , And the upper limit of the closed residual.
[0145] In summary, this invention belongs to the technical fields of digital current control of permanent magnet synchronous motors, space vector pulse width modulation, and electronic speed controllers for unmanned aerial vehicles. It proposes a method and system for calculating the switching events that have actually been executed within the current carrier sub-cycle as prefix voltage action quantities according to the motor's precise state transition exponent, and completing robust reachable projection, exponential time edge generation, timer quantization verification, and endpoint current residual feedback within the remaining time.
[0146] This invention samples the stator current at the start of the current control sub-cycle and records the inverter switching states and edge times that have been actually executed from the start of the sub-cycle to the completion of the calculation. Using the same state transition exponent kernel determined by the motor resistance and inductance, the total voltage action required for the target endpoint current and the actual voltage action of the executed prefix are calculated respectively, and the difference between the two is the remaining action gap. The remaining robust reachable domain is formed based on the remaining time, the measured DC bus voltage, the inverter basic voltage vector, and the uncertain action, and the remaining action gap is projected onto its inner executable domain. The action weights of adjacent active vectors and zero vectors are assigned in the exponential time coordinate, and the PWM edge that has not yet occurred in the current sub-cycle is obtained by inverse logarithmic transformation (inverse logarithmic transformation is to solve the physical time from the exponential coordinate value, corresponding to the natural logarithmic operation). After quantizing the edge to the timer clock, the actual action of the quantized switching sequence is recalculated according to the same exponent kernel, and it is executed only when it meets the robust reachability constraint and the closure residual limit. Otherwise, the feasible edge combination with the smallest closure residual is selected from the adjacent clock combinations. The sampling current in the next sub-cycle is used to generate the physical closure residual and update the uncertain action, thereby forming a closed loop from the executed switching event to the executable edge and then to the measured end current.
[0147] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.
Claims
1. A method for controlling currents of a surface-mounted permanent magnet synchronous motor, characterized in that, Includes the following: At the start of the first current control sub-period The stator current of the surface-mounted permanent magnet synchronous motor is sampled, and the switch event recording unit records the switch event from to the calculation completion time The actual executed inverter switch state, the direct current bus voltage corresponding to each switch state, and the switch edge time According to the stator resistance and the stator inductance, a state transition index core is determined, using the same state transition index core, on one hand, a target total voltage action amount required by a current current control sub-cycle is calculated from a target terminal point current, on the other hand, a prefix voltage action amount already executed is calculated from the inverter switch state actually executed, the DC bus voltage and the switch edge time, and a remaining action amount gap from to the sub-cycle terminal point time is obtained by subtracting the prefix voltage action amount already executed from the target total voltage action amount. According to to the remaining index time measure, the current measured DC bus voltage and the convex hull of the inverter basic voltage vectors, a remaining nominal reachable action domain is constructed, and the remaining nominal reachable action domain is eroded according to the uncertain action formed by the motor parameter error, the back electromotive force action estimation error, the bus voltage error, the dead zone voltage drop and the sampling error, to obtain a remaining robust reachable action domain. projecting the remaining action amount gap to a continuous executable domain located inside the remaining robust reachable action domain to obtain a remaining action amount to be executed; decomposing the remaining action amount to be executed into index time weights of adjacent active voltage vectors and zero voltage vectors in an index time coordinate determined by the state transition index core, and generating a PWM edge time that has not yet occurred The PWM edge time is quantized into a timer clock count value. Based on the quantized actual switching state sequence and the quantized edge time, the quantized actual residual action is recalculated using the same state transition exponent kernel. The timer clock count value is allowed to be loaded only if the quantized actual residual action is within the residual robust reachable action domain and its closure residual relative to the remaining action to be executed is not greater than the preset action residual upper limit. Otherwise, the combination with the smallest closure residual within the residual reachable action domain is selected from adjacent timer clock count combinations. performing the selected PWM edge for the remainder of the current control subcycle, and sampling an end-point stator current, generating a physical closure residual from a difference between the end-point stator current and a predicted end-point current corresponding to the quantized actual remaining effort, and updating the uncertain effort for the next current control subcycle or the inner shrinkage margin of the continuous executable domain using the physical closure residual.
2. The method for controlling the current of a surface-mounted permanent magnet synchronous motor according to claim 1, characterized in that, The surface-mounted permanent magnet synchronous motor in The current model in the stationary coordinate system, the accurate state transition relationship, and the state transition index core are respectively ; (1); ; (2); ;(3); ; (4); wherein, is axis equivalent stator inductance in H; is stator phase resistance in ; , and are the stator current vector, inverter output voltage vector and back EMF vector, respectively, in units of A, V and V, respectively; is the natural decay rate of the current, in s"1; is the natural decay coefficient of the current for one current control sub-period; For the dimensionless state transition index kernel; and V·s, respectively.
3. The method for controlling the current of a surface-mounted permanent magnet synchronous motor according to claim 1, characterized in that, to to the first actual switching state interval , define the exponential time measure and calculate the performed prefix voltage action quantity: ; (5); ; (6); in, This represents the number of actual switch state intervals within the prefix. For the first One actual switch state; The actual voltage vector, in units of V, is obtained by combining the measured DC bus voltage, bridge arm switch status, and dead zone current direction in this section. The unit is s; The unit is V·s.
4. A method for controlling the current of a surface-mounted permanent magnet synchronous motor according to claim 1, characterized in that, The target endpoint current is At that time, the target total voltage action, the residual action gap, and the residual exponential time measure are respectively ; (7); ; (8); The remaining nominally reachable scope, the remaining robustly reachable scope, and the contiguous executable scope are respectively: ; (9); in, This is an estimated value of the back electromotive force, in V·s; The target total voltage effect, in V·s; The remaining action gap is expressed in V·s; Remaining physical time, in seconds; to These are the eight basic voltage vectors of a three-phase two-level inverter, in units of V; This is the set of uncertain action quantities calculated from parameters, back electromotive force, bus, dead zone, and sampling error, in V·s; Reserved action quantity set for timer quantization and minimum pulse width, in V·s; This is the Minkowski set difference operation.
5. A method for controlling the current of a surface-mounted permanent magnet synchronous motor according to claim 1, characterized in that, The remaining action quantity to be executed is determined by the following formula: ; (10); in, This is the positive definite current error weight matrix; when using a contraction line pointing from the current to the target endpoint current, it is defined as follows: ;(11); And will make The largest within the contiguous executable domain Determined as the maximum feasible shrinkage factor ; When the continuous executable domain adopts a hexagonal half-space Indicates and When located within the contiguous executable domain, the following is obtained: ;(12); in, This is the upper limit of the instruction shrinkage factor; For hexagonal number 1 The unit outward normal vector of each face; The shrinkage boundary of the face corresponding to the continuous executable domain, in units of V·s; This is the interval truncation operator.
6. A method for controlling the current of a surface-mounted permanent magnet synchronous motor according to claim 1, characterized in that, The exponential time coordinate, the decomposition of the remaining action to be executed, and the inverse edge transform are respectively: ; (13); ; (14); ; (15); in, The sector number of the voltage vector containing the remaining action to be executed; and These are the two adjacent active voltage vectors of this sector; , and These are the exponential time weights of the zero voltage vector and two adjacent active voltage vectors, respectively, in seconds; A dimensionless exponential time coordinate; The edge time to be quantized is in seconds (s).
7. A method for controlling the current of a surface-mounted permanent magnet synchronous motor according to claim 1, characterized in that, Let the timer clock period be The quantized edge count, quantized edge time, and quantized residual action are respectively... ;(16); ; (17); The conditions for implementing the license are as follows ;(18); in, A quantization operator for selecting timer count values within the continuous executable domain; This represents the number of actual switching state intervals after quantization. The actual residual action after quantification, in V·s; To quantify the closed residual of the executed action relative to the continuous actions to be executed, the unit is V·s; The upper limit of the allowable quantization action error, in V·s; This is the minimum effective pulse width allowed for power devices and drive circuits, measured in seconds (s).
8. A system for controlling the current of a surface-mounted permanent magnet synchronous motor, for performing a method for controlling the current of a surface-mounted permanent magnet synchronous motor as described in any one of claims 1 to 7, characterized in that, It includes a three-phase two-level inverter, a current sampling circuit, a DC bus voltage sampling circuit, a processor, a memory, and a PWM timer, a switching event recording unit, and an intra-carrier edge scheduling unit; wherein the switching event recording unit is used to record the inverter switching state, corresponding DC bus voltage, and edge time that have been actually executed from the start of the current current control sub-cycle to the completion of the calculation; the intra-carrier edge scheduling unit is used to load the remaining PWM edges that have been quantized and verified.
9. A system for controlling the current of a surface-mounted permanent magnet synchronous motor according to claim 8, characterized in that, The surface-mounted permanent magnet synchronous motor is a surface-mounted permanent magnet synchronous motor driven by a three-phase two-level inverter powered by a DC bus.
10. A system for controlling the current of a surface-mounted permanent magnet synchronous motor according to claim 8, characterized in that, It also includes a drone comprising a fuselage, a propeller, a surface-mounted permanent magnet synchronous motor, and an electronic speed controller, wherein the electronic speed controller utilizes the executed prefix voltage action and the remaining robust reachable action domain of the current current control subcycle to generate and execute PWM edges that have not yet occurred in the current current control subcycle without switching current control modes.
