A pwm-based model predictive torque control method

Abstract

The application provides a PWM-based model predictive torque control method, current-flux-position characteristics and torque-current-position characteristics of a switched reluctance motor are obtained through offline measurement and are used to construct a data lookup table, motor phase current and phase torque values are obtained according to current rotor position and motor flux information in combination with the data lookup table, torque at the next moment is estimated, an estimated duty ratio scheme is obtained through LHS design, an RSM equation of torque difference and duty ratio is established, and stationary points of the equation are calculated to obtain optimal duty ratios of input phases and output phases. Simulation verifies the effectiveness of the method, the method has simple logic, good real-time performance and high identification accuracy, and plays an important role in improving the applicability and speed regulation performance of the switched reluctance motor.

Description

Technical Field

[0001] This invention belongs to the field of motor control and relates to a model predictive torque control method based on PWM. Background Technology

[0002] Switched reluctance motors (SRMs) have attracted widespread attention in recent years due to their robust and simple structure, ease of manufacture, high reliability, wide speed range, convenient mode switching, and adaptability to harsh conditions. They have become a strong candidate for drive systems in electric vehicles and multi-electric aircraft.

[0003] Switched reluctance motors (SRMs) are doubly salient pole motors with highly saturated magnetic circuits and nonlinear electromagnetic characteristics, whose output torque cannot be linearly represented by current. However, under traditional current control, the output torque of SRMs fluctuates significantly, and this inherent high torque ripple poses a potential threat to the safety of electric drive systems. To address this issue, researchers have chosen motor control as their research focus, proposing direct torque control (DTC), direct instantaneous torque control (DITC), torque distribution function (TSF), and model predictive control (MPC) to suppress torque ripple in SRMs. Among these, model predictive control's main characteristic is using a system model to predict the future behavior of the controlled variable. Based on predefined optimization criteria, the controller uses this information to obtain the optimal control signal. Therefore, model predictive control has received increasing attention in switched reluctance motor control due to its high efficiency and superior performance.

[0004] However, the applicant found that traditional predictive control uses predictive control signal schemes that only offer three fixed switching states for each phase: excitation state, zero-voltage freewheeling state, and demagnetization state. This method has the following drawbacks:

[0005] 1) The switching frequency is lower, and the waveform obtained at the same calculation frequency is more prone to fluctuation than the waveform obtained by PWM control.

[0006] 2) In model predictive control, this method can obtain the optimal solution among the alternatives, but the calculation results still have a certain gap with the optimal solution.

[0007] Further research by the applicant revealed that Response Surface Modeling (RSM) has been widely applied in motor design optimization. RSM combines mathematical and statistical methods to analyze responses affected by multiple variables, with the ultimate goal of optimizing the response value. Latin Hypercube Sampling (LHS) design divides the domain of each layer into n layers, sampling once from each layer with an equal marginal probability of 1 / n. By using LHS design, the duty cycle varying within [0,1] can be divided into n layers, and a random duty cycle can be obtained from each layer. This not only fills the response surface as much as possible but also ensures the randomness of the duty cycle. Summary of the Invention

[0008] To effectively reduce torque ripple in switched reluctance machines (SRMs), this invention proposes a PWM-based model predictive torque control method for torque ripple suppression. This method consists of deadbeat predictive control (DPC) and model predictive control (MPC). DPC is used to predict the control signal for single-phase excitation. In the commutation region, MPC and a response surface model (RSM) are combined to predict the control signals for both the input and output phases. Experiments under different conditions compare this predictive control method with other typical control methods, and the experimental results demonstrate its superiority.

[0009] The technical solution of this invention is as follows:

[0010] A PWM-based model predictive torque control method includes the following steps:

[0011] Step 1: Obtain the switched reluctance current characteristics, torque characteristics, motor flux linkage characteristics, and rotor position characteristics, and construct data table i. ph (Ψ ph ,θ) and T ph (i ph ,θ); where i ph For phase current, T ph For phase torque, Ψ ph θ is the phase flux linkage value, and θ is the rotor position;

[0012] Step 2: Determine whether the motor is in the single-phase or commutation region. If it is in the single-phase region, use DPC to predict the control signal for the single-phase excitation; otherwise, the control method is determined by the proposed MPC method, i.e., proceed with steps 3-7.

[0013] Step 3: Assume that the flux linkages of each phase at time k+1 are consistent with those at time k, and based on the data table i obtained in Step 1... ph (Ψ ph ,θ) and T ph (i ph ,θ) to obtain the phase current and phase torque, that is, to obtain the predicted phase torque T ph (k+1), the three-phase torques are added together to obtain the predicted output torque T0(k+1);

[0014] Step 4: Compare T0(k+1) with the reference torque T ref The predicted duty cycle scheme for the input and output phases is obtained: if T ref If T0(k+1), then the predicted duty cycle D of the excitation phase is... in The scheme is positive, and the output phase D is positive. out The duty cycle scheme can be in the range of [-1, 1]; if T ref<T0(k+1), the prediction scheme for the output phase should be a negative value, and the duty ratio range of the prediction scheme for the excitation phase can be [-1,1];

[0015] Step 5: For all input and output duty ratios in the prediction scheme, via

[0016] ψ ph (k+1)=ψ ph (k)+[U dc D p -i ph (k)R]T s

[0017] respectively predict the flux linkage Ψ of the input phase and output phase at time k+1 ph (k+1), wherein Ψ ph is the phase flux linkage value, and D p is a corresponding duty ratio.

[0018] Obtain the predicted input and output torque T at time k+1 by looking up tables i ph (Ψ ph ,θ) and T ph (i ph ,θ) at time k+1 in (k+1) and T out (k+1). Then, the absolute value of the difference between the reference torque and the predicted output torque is calculated by the following formula:

[0019] △T=|T ref -T in (k+1)-T out (k+1)|

[0020] Step 6: Establish an RSM equation of torque difference and duty ratio.

[0021]

[0022] Wherein, a1-a6 are regression coefficients, D in and D out are respectively the duty ratios of the input phase and the output phase. To evaluate the accuracy of the equation, the multiple correlation coefficient R 2 is introduced, which is defined as:

[0023]

[0024] wherein is the average experimental data, and is the predicted value of RSM. When the R 2 coefficient is greater than 0.9, the fitting equation can be considered sufficiently accurate.

[0025] After establishing the RSM equations, find the stationary points of the equations:

[0026]

[0027]

[0028] By further calculating the partial derivatives, it can be determined whether the stationary point is a minimum point.

[0029]

[0030]

[0031]

[0032] R = AC - B 2

[0033] If both A and R are positive, then the stationary point is a minimum point on the response surface, and the corresponding duty cycles D1 and D2 will be the optimal duty cycles for the input and output phases.

[0034] If the fitting accuracy is insufficient (R) 2 If the value is <0.9) or there is no minimum point, the scheme that minimizes ΔT can be found in the prediction duty cycle scheme, and the corresponding duty cycle is used as the output.

[0035] Step 7: Convert the optimal duty cycle into a control signal and apply it to the SRM drive system.

[0036] Furthermore, the steps for determining the single-phase and commutation regions of the motor in step 2 are as follows:

[0037] First, calculate the upper limit of the demagnetization time for the output phase:

[0038]

[0039] Δψ represents the change in flux linkage of the output phase during the demagnetization time, U dc This is the bus voltage.

[0040] The time t from the current rotor position θ to the end of the motor torque segment c The calculation is as follows:

[0041]

[0042] t dmax With t c Compare, if t dmax >t c If the signal is in the single-phase region, DPC is used to predict the control signal for the single-phase excitation; otherwise, the control method is MPC.

[0043] Furthermore, the steps in step 2 for predicting the control signal for single-phase excitation using DPC are as follows:

[0044] Predict the flux linkage at time k+1 in the two demagnetizing phases respectively:

[0045] ψ ph (k+1)=ψ ph (k)+[U dc D ph -i ph (k)R]T s

[0046] Among them, U dc D is the bus voltage. p =-1 is the control signal for the two demagnetizing phases, i ph R is the phase current, R is the phase resistance, and T is the phase resistance. s The sampling period.

[0047] Predict the current rotor position:

[0048] θ(k+1)=θ(k)+ωT s

[0049] From the data table i in step 1 ph (Ψ ph ,θ) and T ph (i ph The predicted torque T for the two demagnetizing phases is obtained by (θ). d1 and T d2 The reference torque T of the excitation phase eref Represented as:

[0050] T eref =T ref -T d1 (k+1)-T d2 (k+1)

[0051] According to the calculated T eref Using the data table i from step 1 ph (Ψ ph ,θ) and T ph (i ph The predicted current i of the excitation phase is obtained by θ). ph (k+1) and magnetic flux ψ ph (k+1). Calculate the required duty cycle D. e (k+1):

[0052]

[0053] Furthermore, in step 4, the predicted duty cycle scheme can be obtained in the following way:

[0054] If Tref >T0(k+1), the current output torque is insufficient to meet the reference torque, so it is necessary to increase the output torque by increasing the flux linkage. When the motor is in the commutation region, two-phase control signals need to be calculated, and the third phase can be determined to be demagnetized because it cannot output motoring torque. Meanwhile, it is expected that the incoming phase can replace the outgoing phase to provide torque, therefore, it can be determined that the flux linkage of the incoming phase needs to be increased in the next cycle, and the average voltage applied in the next cycle of the incoming phase must be positive. The flux linkage of the outgoing phase is in an uncertain state, and the average voltage applied in the next cycle can be either positive or negative.

[0055] If T ref <T0(k+1), the flux linkage needs to be reduced. Therefore, it can be determined that the flux linkage of the outgoing phase needs to be reduced in the next cycle, and the average voltage applied in the next cycle is negative, while the average voltage applied in the next cycle of the incoming phase can be positive or negative.

[0056] Further, in step 6, the regression coefficients a1-a6 can be calculated by the least square method, and the method is as follows:

[0057] α=(|D in D out | T ·|D in D out |) -1 ·|D in D out | T ·△T

[0058] where α represents a matrix formed by the regression coefficients a1-a6.

[0059] Further, the method for converting the duty cycle into a PWM control signal in step 7 is as follows:

[0060] A method for modulating a PWM wave by a single-phase asymmetric half-bridge is adopted. When modulating the PWM wave, a triangular carrier varying from -1 to 1 is selected as the modulation carrier. Comparison values c1 and c2 are set to be compared with the triangular carrier to generate two switching signals for switching tubes s1 and s2, and the quantitative relationship between the comparison values c1, c2 and the duty cycle D is as follows:

[0061]

[0062] The generation logic of the s1 control signal is: output a high level when the triangular carrier value is less than the comparison value c1, and output a low level when the carrier value is greater than the comparison value c1; the generation logic of the s2 control signal is: output a low level when the carrier value is less than the comparison value c2, and output a high level when the carrier value is greater than the comparison value c2. When the duty cycle is greater than 0, the phase voltage is between 0 and U dcThe phase voltage fluctuates between 0 and -U when the duty cycle is less than 0. dc It fluctuates between [variables]. Therefore, within a control cycle T... s Within this range, the average voltage applied to a single phase is [-U dc U dc ].

[0063] Beneficial effects

[0064] This invention combines MPC and RSM to construct a response surface model of torque difference and duty cycle, predicting the optimal duty cycle in the commutation region. It proposes a PWM-based model predictive torque control method for torque ripple suppression in switched reluctance motors (SRMs), providing a new approach to obtaining the optimal control signal. Furthermore, this method employs LHS design to obtain the predicted duty cycle scheme, which not only fills the response surface as much as possible but also increases the degrees of freedom of the PWM control signal. By using PWM as the control signal, a control signal at least as high as the control frequency can be output at the same computation frequency, and the prediction scheme can be diversified, making the output control signal closer to the optimal solution.

[0065] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0066] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0067] Figure 1 This is a system block diagram of a PWM-based model predictive torque control method.

[0068] Figure 2 A flowchart illustrating the steps of the improved MPC method.

[0069] Figure 3 This is a schematic diagram of converting the duty cycle into a PWM control signal.

[0070] Figure 4 , Figure 5 and Figure 6 Simulation diagrams are shown for using MATLAB / Simulink software to simulate the overall model prediction system, the DPC module, and the MPC module.

[0071] Figure 7 and Figure 8 The results show the measured phase current, phase torque, and total torque at 700 rpm (2 Nm) and 1000 rpm (1 Nm) for the traditional MPC and the proposed method, respectively. Figure 7 It is 700 rpm (2 Nm). Figure 8 It is 1000 rpm (1 Nm). Detailed Implementation

[0072] To effectively reduce torque ripple in switched reluctance motors (SRMs), this invention proposes a PWM-based model predictive torque control method that predicts the optimal duty cycle in the commutation region for torque ripple suppression in SRMs. The embodiments of this invention are described in detail below. These embodiments are exemplary and intended to explain the invention, and should not be construed as limiting the invention.

[0073] The proposed composite torque model predictive control method was tested on a three-phase 12 / 8-pole SRM electric drive system, with the traditional MPC as a comparison, and the torque ripple suppression effect was verified. Figure 7 and Figure 8 The measured results of phase current, phase torque, and total torque at 700 rpm (2 Nm) and 1000 rpm (1 Nm) are shown under conventional MPC and the proposed prediction method.

[0074] Step 1: Obtain the switched reluctance current characteristics, torque characteristics, motor flux linkage characteristics, and rotor position characteristics, and construct data table i. ph (Ψ ph ,θ) and T ph (i ph ,θ); where i ph For phase current, T ph For phase torque, Ψ ph θ is the phase flux linkage value, and θ is the rotor position;

[0075] Step 2: Determine the single-phase and commutation regions of the motor.

[0076] First, calculate the upper limit of the demagnetization time for the output phase:

[0077]

[0078] Δψ represents the change in flux linkage of the output phase during the demagnetization time, U dc This is the bus voltage.

[0079] The time t from the current rotor position θ to the end of the motor torque segment c The calculation is as follows:

[0080]

[0081] t dmax With t c Compare, if t dmax >t c If the signal is in the single-phase region, DPC is used to predict the control signal for the single-phase excitation; otherwise, the control method is determined by MPC, i.e., steps 3-7 are taken.

[0082] Step 3: Assume that the flux linkages of each phase at time k+1 are consistent with those at time k, and obtain the data table i obtained in step 1 ph (Ψ ph ,θ) and T ph (i ph ,θ) to obtain the phase current and phase torque, that is, obtain the predicted phase torque T ph (k+1), and adding the three-phase torques obtains the initial predicted output torque T0(k+1);

[0083] Step 4: Compare T0(k+1) with the reference torque T ref to obtain the predicted duty cycle schemes for the magnetizing phase and the demagnetizing phase. If T ref >T0(k+1), then the predicted duty cycle D of the magnetizing phase in scheme is positive, and the duty cycle scheme of output phase D out varies within [-1,1]; if T ref <T0(k+1), then the predicted scheme of the demagnetizing phase is negative, and the predicted scheme of the magnetizing phase varies within [-1,1];

[0084] Step 5: For all input and output duty cycles in the prediction scheme, through

[0085] ψ ph (k+1)=ψ ph (k)+[U dc D ph -i ph (k)R]T s

[0086] the flux linkages Ψ of the magnetizing phase and the demagnetizing phase at time k+1 are predicted respectively ph (k+1).

[0087] By looking up table i ph (Ψ ph ,θ) and T ph (i ph ,θ), the predicted input torque and output torque T at time k+1 are obtained in (k+1) and T out (k+1). Then, calculate the absolute value of the difference between the reference torque and the predicted output torque by the following formula:

[0088] △T=|T ref -T in (k+1)-T out (k+1)|

[0089] where Ψ ph is the phase flux linkage value, U dc is the voltage value, Dph This represents the corresponding duty cycle.

[0090] Step 6: Establish the RSM equations for torque difference and duty cycle.

[0091]

[0092] Where a1-a6 are regression coefficients, D in and D out These are the duty cycles of the excitation phase and the demagnetization phase, respectively.

[0093] Calculate the multiple correlation coefficient R 2 Determine the fitting accuracy.

[0094]

[0095] in These are average experimental data. This is the predicted value of RSM.

[0096] Calculate the stationary points of the equation:

[0097]

[0098]

[0099] Calculate the partial derivatives to determine whether the stationary point is a local minimum:

[0100]

[0101]

[0102]

[0103] R = AC - B 2

[0104] If both A and R are positive, then the stationary point is a minimum point on the response surface, and the corresponding duty cycles D1 and D2 will be the optimal duty cycles for the excitation and demagnetization phases.

[0105] If the fitting accuracy is insufficient (R) 2 If the value is less than 0.9 or there is no minimum point, find the scheme that minimizes ΔT in the predicted duty cycle scheme and output the corresponding duty cycle.

[0106] Step 7: The optimal duty cycle is converted into a control signal by using a single-phase asymmetric half-bridge delta carrier modulation PWM wave and applied to the SRM drive system.

[0107] It can be seen that this method, after discarding the excitation angle and torque distribution function, can still perform torque control in the commutation zone of the SRM, which suppresses torque pulsation and prevents the occurrence of negative torque.

[0108] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention without departing from the principles and spirit of the present invention.

Claims

1. A model predictive torque control method based on PWM, characterized in that: Includes the following steps: Step 1: Obtain the switched reluctance current characteristics, torque characteristics, motor flux linkage characteristics, and rotor position characteristics, and construct data table i. ph (Ψ ph ,θ) and T ph (i ph ,θ); where i ph For phase current, T ph For phase torque, Ψ ph θ is the phase flux linkage value, and θ is the rotor position; Step 2: Determine whether the motor is in the single-phase or commutation region; if it is in the single-phase region, use DPC to predict the control signal of the single-phase excitation; otherwise, the control method is determined by the proposed MPC method, i.e., steps 3-7 are taken. Step 3: Assume that the flux linkages of each phase at time k+1 are consistent with those at time k, and based on the data table i obtained in Step 1... ph (Ψ ph ,θ) and T ph (i ph ,θ) to obtain the phase current and phase torque, that is, to obtain the predicted phase torque T ph (k+1), the three-phase torques are added together to obtain the predicted output torque T0(k+1); Step 4: Compare T0(k+1) with the reference torque T ref to obtain a predicted duty cycle scheme for the input and output phases: if T ref > T0(k+1), then the predicted duty ratio D of the excitation phase in scheme is positive, and the duty ratio scheme range of output phase D out can be [-1,1]; if T ref < T0(k+1), then the prediction scheme of the output phase should be negative, and the duty ratio range of the prediction scheme of the excitation phase can be [-1,1]; Step 5: For all input and output duty cycles in the prediction scheme, by... ψ ph (k+1)=ψ ph (k)+[U dc D p -i ph (k)R]T s Predict the flux linkage Ψ for the input and output phases at time k+1, respectively. ph (k+1), where D p For the corresponding duty cycle; By looking up table i ph (Ψ ph ,θ) and T ph (i ph The predicted input and output torques T at time k+1 were obtained (θ). in (k+1) and T out (k+1); then, the absolute value of the difference between the reference torque and the predicted output torque is calculated using the following formula: △T=T ref -T in (k+1)-T out (k+1) Step 6: Establish the RSM equations for torque difference and duty cycle; Where a1-a6 are regression coefficients, D in and D out These are the duty cycles of the input and output phases, respectively; to evaluate the accuracy of the equation, a complex correlation coefficient R is introduced. 2 Its definition is: in These are average experimental data. It is the predicted value of RSM; when R 2 When the coefficient is greater than 0.9, the fitted equation can be considered to be sufficiently accurate; After establishing the RSM equations, find the stationary points of the equations: By further calculating the partial derivatives, it can be determined whether the stationary point is a minimum point; R=AC-B 2 If both A and R are positive, then the stationary point is a minimum point on the response surface, and the corresponding duty cycles D1 and D2 will be the optimal duty cycles for the input and output phases. If the fitting accuracy is insufficient, i.e., R 2 If the value is less than 0.9 or there is no local minimum, the scheme that minimizes ΔT can be found in the predicted duty cycle scheme, and the corresponding duty cycle can be used as the output. Step 7: Convert the optimal duty cycle into a control signal and apply it to the SRM drive system.

2. The PWM-based model predictive torque control method according to claim 1, characterized in that: The steps for determining the single-phase and commutation regions of the motor in step 2 are as follows: First, calculate the upper limit of the demagnetization time for the output phase: Δψ represents the change in flux linkage of the output phase during the demagnetization time, U dc This refers to the bus voltage. The time t from the current rotor position θ to the end of the motor torque segment c The calculation is as follows: t dmax With t c Compare, if t dmax >t c If the signal is in the single-phase region, DPC is used to predict the control signal for the single-phase excitation; otherwise, the control method is MPC.

3. The PWM-based model predictive torque control method according to claim 1, characterized in that: The steps for predicting the control signal for a single-phase excitation using DPC in step 2 are as follows: Predict the flux linkage at time k+1 in the two demagnetizing phases respectively: ψ ph (k+1)=ψ ph (k)+[U dc D ph -i ph (k)R]T s Among them, U dc D is the bus voltage. p =-1 is the control signal for the two demagnetizing phases, i ph R is the phase current, R is the phase resistance, and T is the phase resistance. s The sampling period; Predict the current rotor position: θ(k+1)=θ(k)+ωT s From the data table i in step 1 ph (Ψ ph ,θ) and T ph (i ph The predicted torque T for the two demagnetizing phases is obtained by (θ). d1 and T d2 The reference torque T of the excitation phase eref Represented as: T eref =T ref -T d1 (k+1)-T d2 (k+1) According to the calculated T eref Using the data table i from step 1 ph (Ψ ph ,θ) and T ph (i ph The predicted current i of the excitation phase is obtained by θ). ph (k+1) and magnetic flux ψ ph (k+1); Calculate the required duty cycle D e (k+1):

4. The PWM-based model predictive torque control method according to claim 1, characterized in that: In step 4, the predicted duty cycle scheme can be obtained in the following way: If T ref If the current output torque is less than T0(k+1), then the output torque is insufficient to meet the reference torque, and the output torque needs to be increased by increasing the flux linkage. The motor is in the commutation zone, and two-phase control signals need to be calculated. The third phase can be determined to be demagnetized because it cannot output motor torque. At the same time, we expect the input phase to replace the output phase to provide torque. Therefore, it can be determined that the flux linkage of the input phase needs to be increased in the next cycle, and the average voltage applied to the input phase in the next cycle must be positive. The flux linkage of the output phase is in an uncertain state, and the average voltage applied in the next cycle can be positive or negative. If T ref <T0(k+1), the flux linkage needs to be reduced; Therefore, it can be determined that the flux linkage of the output phase needs to be reduced in the next cycle, and the average voltage applied in the next cycle is negative, while the average voltage applied in the next cycle of the input phase can be positive or negative.

5. The PWM-based model predictive torque control method according to claim 1, characterized in that: In step 6, the regression coefficients a1-a6 can be calculated using the least squares method, as follows: α=(D in D out T ·D in D out ) -1 ·D in D out T ·△T Where α represents the matrix formed by regression coefficients a1-a6.

6. The PWM-based model predictive torque control method according to claim 1, characterized in that: The method for converting the optimal duty cycle into a PWM control signal in step 7 is as follows: A single-phase asymmetric half-bridge modulation PWM wave method is adopted. When modulating the PWM wave, a triangular carrier wave varying from -1 to 1 is selected as the modulation carrier. Comparison values ​​c1 and c2 are set to be compared with the triangular carrier wave to generate two switching signals for switching transistors s1 and s2. The quantitative relationship between comparison values ​​c1 and c2 and the duty cycle D is as follows: The logic for generating the S1 control signal is as follows: when the triangular carrier value is less than the comparison value c1, the output is high; when the carrier value is greater than the comparison value c1, the output is low. The logic for generating the S2 control signal is as follows: when the carrier value is less than the comparison value c2, the output is low; when the carrier value is greater than the comparison value c2, the output is high. When the duty cycle is greater than 0, the phase voltage is between 0 and U. dc The phase voltage fluctuates between 0 and -U when the duty cycle is less than 0. dc Jumping between; Therefore, in a control cycle T s Within this range, the average voltage applied to a single phase is [-U dc U dc ].

Images