A model prediction based fixed switching frequency motor drive modulation method

Abstract

This invention discloses a model-predictive fixed-frequency motor drive modulation method. It relates to the fields of power electronics and electric drive control technology, and includes: obtaining the clamped duty cycle based on the current motor state; constructing an extended switch sequence set based on this, using a single variable to represent the pulse initial conduction time, deriving the relationship between the sub-interval action time and this variable, and the feasible region; calculating the current gradient of each sub-interval, recursively predicting the switching point current, and obtaining the state vector; constructing a cost function in the form of the squared L2 norm of the difference between the state vector and the reference vector, solving for the unconstrained optimal solution and projecting it onto the feasible region to obtain the optimal initial time of each sequence; merging the traditional SVPWM sequence and the extended switch sequence into a unified candidate set, constructing a unified cost function to calculate the minimum cost value of each sequence; selecting the sequence with the minimum cost value and its optimal solution to generate an augmented duty cycle vector, which is then converted into a gate signal to drive the inverter via dual-threshold pulse width modulation.

Description

Technical Field

[0001] This invention relates to the field of power electronics and electric drive control technology, and more specifically to a fixed switching frequency motor drive modulation method based on model prediction. Background Technology

[0002] With increasing demands on the performance of motor drive systems, MPCs have gained widespread attention due to their fast dynamic response and strong constraint handling capabilities. However, traditional FCS-MPCs typically generate variable switching frequencies, leading to wide-spectrum current harmonics, which increases the difficulty of filter design and motor losses.

[0003] On the other hand, although modulator-based field-oriented control combined with space vector pulse width modulation or discontinuous pulse width modulation (DPWM) can maintain a fixed switching frequency, its current harmonic performance will significantly decrease under high speed (high modulation ratio) or switching frequency limited (low carrier ratio) conditions, resulting in obvious baseband distortion and sideband expansion.

[0004] Traditional pulse width modulation (PWM): Under high modulation ratio and low carrier ratio conditions, the fixed switching sequence and action time calculation method make it impossible to flexibly adjust the pulse distribution, resulting in obvious low-order harmonics (such as widening of sideband harmonic clusters and destruction of half-wave symmetry leading to even-order harmonics), which causes a significant increase in current THD.

[0005] Traditional finite control set model predictive control (FCS-MPC) has a non-fixed switching frequency, a dispersed spectrum, and can only select the basic voltage vector at each moment, resulting in large current ripple and affecting steady-state performance.

[0006] Optimal Switch Sequence Predictive Control (OSS-MPC): The candidate sequence set is limited, and a complex multidimensional optimization problem needs to be solved.

[0007] Optimal Pulse Mode (OPP): Relies on a large offline lookup table, making it complex to implement and difficult to implement online in low-cost controllers.

[0008] Therefore, how to provide a model prediction method for switching sequence selection with a fixed switching frequency, which can significantly reduce current harmonic distortion under high modulation ratio or low carrier ratio conditions, and has low computational complexity and is easy to implement in engineering, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0009] In view of the above problems, the present invention is proposed to provide a model prediction-based fixed switching frequency motor drive modulation method that overcomes or at least partially solves the above problems.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, embodiments of the present invention provide a fixed-frequency motor drive modulation method based on model prediction, characterized in that it includes: S1. Based on the motor status and reference voltage obtained in the current control cycle, calculate the three-phase duty cycle, identify the phase with the smallest duty cycle, and inject common-mode voltage to clamp the phase to one end of the DC bus to obtain the clamped duty cycle vector. S2. Construct an extended switch sequence based on the clamped duty cycle vector. gather Extended switch sequence This includes allocating the number of released switches to one of the two unclamped phases, causing that phase to generate a double pulse within one control cycle, and using a single variable. This represents the initial conduction time of the phase pulse. Under the constraints of duty cycle conservation and centrosymmetry, the action time of each sub-interval is derived with respect to... Linear or affine relation and The corresponding feasible region; S3. Calculate the current gradient of each sub-interval based on the current current state and the voltage vector applied to each sub-interval. By combining the sub-interval action time, the current value of each switching point is recursively predicted, and the switching point state vector is obtained. ; S4. Construct a state vector for the switching point. The cost function is of the form of the squared L2 norm of the difference from the reference current vector. The cost function is A quadratic function in one variable is used to find the unconstrained optimal value based on the cost function. The optimal solutions for the initial time steps of each switching sequence are obtained by projecting them onto the feasible region. ; S5. Assemble traditional Space Vector Pulse Width Modulation (SVPWM) sequences. With extended switch sequence gather Merged into a unified candidate set Construct a unified cost function for evaluation. Calculate the minimum cost of each candidate sequence. ; S6. Selecting the Value minimal sequence Its corresponding optimal solution Generate augmented duty cycle vector And use dual-threshold pulse width modulation to augment the duty cycle vector It is converted into a gate signal to drive the inverter.

[0011] Furthermore, in S2, under the constraints of duty cycle conservation and central symmetry, the action time of each sub-interval is derived with respect to... The specific process of establishing the linear or affine relation and its feasible region is as follows: Introducing general variables Indicates the duty cycle of the unclamped two phases and general variables. To represent the duty cycle of the clamping phase, the sub-interval action time vector is defined as... , right It exhibits a linear or affine relationship, and its specific expression is:

[0012] The corresponding feasible region is: .

[0013] in For a complete control cycle, 1，...，6 This represents a switch sequence.

[0014] Furthermore, in S3, the current gradient of each sub-interval is calculated. The specific process is as follows: Based on the continuous model of a permanent magnet synchronous motor in the dq system, the current gradient of each sub-interval is approximated as constant within one period. For the ... Sub-intervals, current gradient for:

[0015] in This indicates the current current state. , , These are three coefficient matrices composed of motor parameters and operating conditions; This is the coordinate transformation matrix; For the first The voltage vector corresponding to the switching state applied to each sub-interval.

[0016] Furthermore, the switching point state vector is obtained in S3. The specific process is as follows: Recursively predict the current at each switching point:

[0017] in For extended switch sequences The current vector at the j-th switching point; This is an extended switch sequence. The current vector at the (j-1)th switching point Let be the current gradient vector at the j-th switching point. Then it is The duration of action of the corresponding j-th sub-interval; Stack all switching point current values ​​into a switching point state vector , to obtain Affine form:

[0018] in , 6 A unit lower triangular matrix of size 6. It is a second-order identity matrix. in The definition is as follows:

[0019]

[0020]

[0021] in The universal vector represents the duty cycle of the unclamped two phases; For a complete control cycle; For extended switch sequences The corresponding current gradient vectors at the 6 switching points; vector , ,in Let the coefficient vector be defined as: .

[0022] Furthermore, the cost function in S4 The expression is:

[0023] in The reference current vector; Based on cost function Find the unconstrained optimal for:

[0024] Then project onto the feasible region to obtain the optimal initial time solutions for each switching sequence. for: .

[0025] Furthermore, S5 incorporates the traditional Space Vector Pulse Width Modulation (SVPWM) sequence. With extended switch sequence Merged into a unified candidate set for:

[0026] Construct a unified cost function for evaluation. for:

[0027] in The reference current vector; As weight; For augmented duty cycle vector; This is the best result from the previous period.

[0028] Furthermore, in S5, the minimum cost value of each candidate sequence is calculated. The specific process is as follows: If the candidate sequence belongs to the extended switch sequence Then the optimal solution at the initial time can be obtained. and its corresponding minimum cost ; If the candidate sequence belongs to the traditional space vector pulse width modulation (SVPWM) sequence Then the cost is calculated directly from the parsing dwell time.

[0029] Furthermore, the augmented duty cycle vector in S6 The definition of is:

[0030] in The first threshold of the double-pulse phase, The second threshold for the double-pulse phase; It is the single-pulse phase threshold, and =1; The clamping phase threshold is, and All are 0.

[0031] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following: This invention provides a model-predictive-based fixed-switching-frequency motor drive modulation method. Under the constraint of a fixed switching frequency, it significantly improves current quality under high modulation or low carrier ratio conditions and reduces THD and current ripple by extending the switching sequence and optimizing the initial timing. The optimization of multiple dwell times within a cycle is reduced to a single-variable online optimization, resulting in low computational burden and suitability for real-time implementation on DSP / FPGA. Compared to fixed-mode PWM, this invention can adaptively select the traditional or extended sequence based on the prediction error in each cycle and optimize the switching timing, thus exhibiting stronger harmonic shaping capabilities in the deep high modulation region. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0033] Figure 1 This is an overall flowchart provided in the embodiments of the present invention; Figure 2 The present invention provides six combinations of extended switch sequences for clamping the c-phase at the negative end of the DC bus. Figure 3 This is a schematic diagram of the dual-threshold PWM principle provided in an embodiment of the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] This invention discloses a model-predictive-based method for modulating a fixed-switching-frequency motor drive, such as... Figure 1 As shown, it includes: S1. Based on the motor status and reference voltage obtained in the current control cycle, calculate the three-phase duty cycle, identify the phase with the smallest duty cycle, and inject common-mode voltage to clamp the phase to one end of the DC bus to obtain the clamped duty cycle vector. S2. Construct an extended switch sequence based on the clamped duty cycle vector. gather Extended switch sequence This includes allocating the number of released switches to one of the two unclamped phases, causing that phase to generate a double pulse within one control cycle, and using a single variable. This represents the initial conduction time of the phase pulse. Under the constraints of duty cycle conservation and centrosymmetry, the action time of each sub-interval is derived with respect to... Linear or affine relation and The corresponding feasible region; S3. Calculate the current gradient of each sub-interval based on the current current state and the voltage vector applied to each sub-interval. By combining the sub-interval action time, the current value of each switching point is recursively predicted, and the switching point state vector is obtained. ; S4. Construct a state vector for the switching point. The cost function is of the form of the squared L2 norm of the difference from the reference current vector. The cost function is A quadratic function in one variable is used to find the unconstrained optimal value based on the cost function. The optimal solutions for the initial time steps of each switching sequence are obtained by projecting them onto the feasible region. ; S5. Assemble traditional Space Vector Pulse Width Modulation (SVPWM) sequences. With extended switch sequence gather Merged into a unified candidate set Construct a unified cost function for evaluation. Calculate the minimum cost of each candidate sequence. ; S6. Selecting the Value minimal sequence Its corresponding optimal solution Generate augmented duty cycle vector And use dual-threshold pulse width modulation to augment the duty cycle vector It is converted into a gate signal to drive the inverter.

[0036] The specific implementation of this invention is as follows: 1) Construction of extended switching sequence: clamping and redistribution of commutation times (keeping the average switching frequency constant) Control cycle is Traditional three-phase SVPWM involves a fixed number of switching actions in each of the three bridge arms within each cycle, resulting in a fixed average switching frequency. By clamping the phase with the smallest duty cycle (e.g., clamping phase c to the negative terminal of the DC bus), the number of switching actions (commutation) for that phase within a cycle is reduced to 0. This allows for an increase in the number of switching actions for the other two phases (e.g., phase a or phase b). This introduces additional pulses to one phase without changing the overall average switching frequency, forming an extended candidate set of switching sequences. Figure 2 As shown. The mechanism is essentially this: by clamping, it obtains controllable commutation degrees of freedom, and by redistributing the number of switching operations, it transforms these degrees of freedom into more adjustable pulse positions within a cycle, thereby improving current shaping capability.

[0037] 2) Sub-interval action time representation of extended sequences (univariate parameterization) Taking c-phase clamping as an example, an extended switch sequence combination suitable for this method is given. and a vector consisting of the action times of each switch sequence sub-interval. The calculation formula is given first. The extended switch sequence combination is then presented. :

[0038] For each type of extended switch sequence Define its sub-interval action time vector:

[0039] by Taking this as an example and combining it with the extended switch sequence combination To explain vectors ,like Figure 2 As shown in (a), in one control cycle The interface can be divided into seven sub-intervals based on the different switch states: 000, 100, 000, 010, 000, 100, and 000. The sum of the durations of these seven sub-intervals is... (i.e., one control cycle), and because Since the quantities are known, we only need to calculate the duration of action of any six subintervals. Here, we choose to calculate the duration of action of the first six subintervals, denoted as follows: , , , , , The vector formed by the duration of action of these six sub-intervals It is a switch sequence The corresponding sub-interval action time vector.

[0040] Introducing general variables This indicates the duty cycle of the two phases not clamped (the duty cycle after clamping is denoted as...). ) and general variables To represent the duty cycle of the clamping phase (since clamping always has...) ), and use a free variable Indicates the initial turn-on time of the applied pulse phase, such as Figure 2 As shown, Figure 2 In the diagram, (a), (b), and (c) represent the cases where phase a is the applied pulse phase. That is, the moment when phase a is initially turned on within this control cycle. Figure 2 In the diagram, (d), (e), and (f) represent the cases where phase a is the applied pulse phase, and the same applies when phase c is the applied pulse phase. Under the constraints of duty cycle conservation and centrosymmetry, all other switching times of the extended switching sequence are determined by... The only decision, therefore right It exhibits a linear or affine relationship, and its specific expression is as follows: (1) These calculation formulas use general variables. Rather than specific to a particular phase (These represent the duty cycles of phases a, b, and c, respectively), ensuring that the above formula can be used to solve for the sub-interval duration vector regardless of which phase is clamped. The mapping relationship is as follows: when phase c is clamped, its duty cycle... When phase a is clamped, its duty cycle When phase b is clamped, its duty cycle .

[0041] Since the action time of the subinterval is nonnegative, we can obtain The corresponding feasible region: (2).

[0042] 3) Predictive Model and Cost Function: Transforming multi-mode evaluation into univariate quadratic optimization. Based on the continuous model of a permanent magnet synchronous motor (PMSM) in the dq coordinate system, the current gradient of each sub-interval is approximated as constant within one cycle. For the ... Sub-intervals, current gradient for: (3) in , The switching state (corresponding voltage vector) applied to this sub-interval. , , These are three coefficient matrices composed of motor parameters and operating conditions. The coordinate transformation correlation matrix:

[0043]

[0044]

[0045]

[0046]

[0047]

[0048] This is the DC bus voltage. For PMSM stator resistors, These are direct-axis and quadrature-axis inductors, respectively. It is electric angular velocity. It is the rotor electrical angle. It is a magnetic flux.

[0049] Recursively predict the current at each switching point: (4) in For extended switch sequences The current vector at the j-th switching point This is the current vector at the (j-1)th switching point. Let be the current gradient vector at the j-th switching point. Then it is The duration of the corresponding j-th sub-interval. Stack the six switching point states into a switching point state vector. , can obtain the Affine form: (5) in, , 6 A unit lower triangular matrix of size 6. It is a second-order identity matrix. ( The definition is as follows:

[0050]

[0051]

[0052] And switch sequence The current gradient vectors corresponding to the 6 switching points :

[0053] ( ) is the first Current gradient at each switching point; vector , coefficient vector From the initial turn-on time of the applied pulse phase In the switch sequence Corresponding sub-interval action time vector The coefficients in expression (1) are defined as follows:

[0054] Construct cost function : (6) in The reference current vector. Because right It's an affine. yes Given a quadratic function in one variable, we can first find the unconstrained optimal solution. : (7) Then project it onto the feasible region to obtain the optimal solution for the initial time corresponding to each switching sequence. : (8) 4) Unified comparison and final selection of SVPWM and extended switching sequence For each candidate sequence: like : Follow the steps above to find Substitute it into formula (1) to calculate. The value is determined based on the start time of the sub-interval where the switch state is 1 in the applied pulse phase. Finally, calculate the cost. .by Figure 2 (a) and its corresponding switching sequence For example: for a switch sequence The unconstrained optimal solution is obtained by successively applying formulas (7) and (8). and the optimal solution at the initial time The initial optimal solution obtained Substituting into formula (1), we can obtain the switching sequence. The corresponding subinterval action vector After obtaining The end time of the first pulse was then calculated as follows: , and then combine Figure 3 (a) It can be known ( From 0 to The slope of the triangular wave within a given time period), and similarly... And for this time Since it is a single pulse and the triangular carrier has symmetry, only the pulse start time needs to be calculated. , and then combine Figure 3 (b) Calculate , Phase C is clamped at the negative end of the DC bus. This concludes the process. The calculation. Calculate Then substitute it into formula (10) to calculate. That's all.

[0055] like The dwell time is directly obtained from the SVPWM analytical expression and requires no optimization.

[0056] To ensure a unified comparison, the traditional SVPWM sequence set is used. With extended sequence set merge: (9) Construct a unified cost function for evaluation: (10) This cost function can be used to evaluate current tracking error and, to some extent, suppress switching of the switching sequence mode during the cycle. To augment the duty cycle vector, This is the best result from the previous period. For weights.

[0057] Final choice: (11) in This represents the optimal switch sequence corresponding to the minimum cost function value. This indicates that the sequence is in the merged switch sequence. The serial number or index value in the data. Indicates the merged The corresponding switch sequence, express The corresponding sequence is substituted into the calculated cost value.

[0058] according to Optimal augmented duty cycle corresponding to different outputs ,when hour, ;when hour, It can be obtained directly from the SVPWM analytical expression.

[0059] 5) Augmented duty cycle and dual-threshold PWM implementation (one-cycle double pulse) To support flexible pulse positions within a cycle (e.g., implementing dual pulses in a certain phase), this invention employs dual threshold comparison logic. For each phase... The augmented duty cycle vector is defined as:

[0060] When the carrier value is in the range Inner Season ,otherwise Its constraints satisfy (Required duty cycle for this phase). When taking It degenerates into a traditional single-threshold single-pulse. This dual-threshold PWM is naturally matched with the "add pulse" structure of the extended sequence, enabling richer pulse orchestration.

[0061] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0062] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fixed-frequency motor drive modulation method based on model prediction, characterized in that, include: S1. Based on the motor status and reference voltage obtained in the current control cycle, calculate the three-phase duty cycle, identify the phase with the smallest duty cycle, and inject common-mode voltage to clamp the phase to one end of the DC bus to obtain the clamped duty cycle vector. S2. Based on the clamped duty cycle vector, construct an extended switch sequence. gather The extended switch sequence This includes allocating the number of released switches to one of the two unclamped phases, causing that phase to generate a double pulse within one control cycle, and using a single variable. This represents the initial conduction time of the phase pulse. Under the constraints of duty cycle conservation and centrosymmetry, the action time of each sub-interval is derived with respect to... Linear or affine relation and The corresponding feasible region; S3. Calculate the current gradient of each sub-interval based on the current current state and the voltage vector applied to each sub-interval. By combining the action time of the sub-intervals, the current value of each switching point is recursively predicted to obtain the switching point state vector. ; S4. Construct the state vector of the switching point. The cost function is of the form of the squared L2 norm of the difference from the reference current vector. The cost function is A quadratic function in one variable is used to find the unconstrained optimal value based on the cost function. The optimal initial time solution for each switching sequence is obtained by projecting the solution onto the feasible region. ; S5. Assemble traditional Space Vector Pulse Width Modulation (SVPWM) sequences. With extended switch sequence gather Merged into a unified candidate set Construct a unified cost function for evaluation. Traverse the candidate set, perform different operations based on its membership, and calculate the minimum cost value for each candidate sequence. ; S6. Select the aforementioned cost value minimal sequence Its corresponding optimal solution Generate augmented duty cycle vector The augmented duty cycle vector is then modulated using dual-threshold pulse width modulation. It is converted into a gate signal to drive the inverter.

2. The method as described in claim 1, characterized in that, In S2, under the constraints of duty cycle conservation and central symmetry, the action time of each sub-interval is derived as follows: The specific process of establishing the linear or affine relation and its feasible region is as follows: Introducing general variables Indicates the duty cycle of the unclamped two phases and general variables. The duty cycle of the clamping phase is represented by the sub-interval action time vector, which is defined as follows: in , right It exhibits a linear or affine relationship, and its specific expression is: The corresponding feasible region is: in For a complete control cycle, 1，...，6 This represents a switch sequence.

3. The method as described in claim 1, characterized in that, The current gradient of each sub-interval is calculated in S3. The specific process is as follows: Based on the continuous model of the permanent magnet synchronous motor in the dq system, the current gradient of each sub-interval is approximated as constant within one period. For the th Sub-intervals, current gradient for: in This indicates the current current state. , , These are three coefficient matrices composed of motor parameters and operating conditions; This is the coordinate transformation matrix; For the first The voltage vector corresponding to the switching state applied to each sub-interval.

4. The method as described in claim 1, characterized in that, The switching point state vector is obtained in S3. The specific process is as follows: Recursively predict the current at each switching point: in For the extended switch sequence The current vector at the j-th switching point; Then it is the extended switch sequence The current vector at the (j-1)th switching point Let be the current gradient vector at the j-th switching point. Then it is The duration of action of the corresponding j-th sub-interval; Stack all switching point current values ​​into a switching point state vector , to obtain Affine form: in , 6 A unit lower triangular matrix of size 6. A second-order identity matrix in The definition is as follows: in The universal vector represents the duty cycle of the unclamped two phases. For a complete control cycle; For extended switch sequences The corresponding current gradient vectors at the 6 switching points; vector , ,in Let the coefficient vector be defined as: .

5. The method as described in claim 4, characterized in that, The cost function in S4 The expression is: in The reference current vector; Based on the cost function Find the unconstrained optimal for: Then project onto the feasible region to obtain the optimal initial time solutions for each switching sequence. for: in t max Indicates the upper limit of the feasible region. t min This represents the lower bound of the feasible region.

6. The method as described in claim 1, characterized in that, In S5, the traditional Space Vector Pulse Width Modulation (SVPWM) sequence is used. With extended switch sequence Merged into a unified candidate set for: Construct a unified cost function for evaluation. for: in The reference current vector; As weight; For augmented duty cycle vector; This is the best result from the previous period.

7. The method as described in claim 1, characterized in that, In step S5, the minimum cost of each candidate sequence is calculated. The specific process is as follows: If the candidate sequence belongs to the extended switch sequence Then the optimal solution at the initial time can be obtained. and its corresponding minimum cost ; If the candidate sequence belongs to the traditional space vector pulse width modulation (SVPWM) sequence Then the cost is calculated directly from the parsing dwell time.

8. The method as described in claim 1, characterized in that, The augmented duty cycle vector in S6 The definition of is: in The first threshold of the double-pulse phase, The second threshold for the double-pulse phase; It is the single-pulse phase threshold, and =1; The clamping phase threshold is, and All are 0.

Images