An MPC-based method for suppressing oscillations in hybrid transmission mode switching by active motor response

Abstract

The present invention discloses a motor active response hybrid transmission mode switching oscillation suppression method based on MPC, which is suitable for the rapid following of the motor speed and rapid power compensation of hybrid electric vehicles. An active response oscillation suppressor is established, and the rapid response of the motor speed following is achieved through the MPC control strategy. The prediction control module and the torque compensation module in the oscillation suppressor cooperate with each other to finally output a set of reference voltages to the modulator. The oscillation during the mode switching of the hybrid vehicle is mainly due to the relative speed of the clutch friction plate. By actively responding the motor speed to the clutch active plate speed, the clutch engagement process is fast and smooth, reducing the system oscillation caused by the motor cutting in. The torque compensation module is used to realize the rapid power compensation of the motor, and the control in the VCU is previously sunk to the MCU, bypassing the communication delay problem, greatly improving the power compensation speed, and realizing the smooth switching of the vehicle mode.

Description

Technical Field

[0001] The invention discloses a method for suppressing oscillation of hybrid transmission mode switching by active response of a motor based on MPC, and belongs to the field of hybrid electric vehicles. Background Art

[0002] Compared with traditional fuel vehicles and pure electric vehicles, hybrid electric vehicles (HEV) have many advantages. The energy source of HEV is fossil fuel, and there is no charging restriction. Therefore, as a bridge from traditional fuel vehicles to pure electric vehicles, HEV has better development prospects than pure electric vehicles in the short term.

[0003] HEV is a comprehensive power system composed of multiple power units. The power required for driving also varies in a large range during actual driving. Therefore, the system will switch between different working modes. The core of this process is the hybrid transmission, which realizes different power configurations by combining power sources in different forms to meet the power requirements corresponding to different working conditions. Therefore, optimizing the control strategy of the hybrid transmission is the focus of improving the performance of HEV. Ensuring the smooth operation of the vehicle during mode conversion is an important goal of hybrid vehicle control, and the stability of mode switching with motor cut-in has become a major focus of studying HEV mode switching strategies. This problem can be expressed as: Can the rapid response of the motor and rapid power compensation after engagement be achieved, thereby suppressing the oscillation caused by mode switching.

[0004] At present, most of the research on HEV mode switching focuses on the switch from pure electric mode to hybrid mode, that is, the impact of engine power intervention on system stability and solutions. The relatively concentrated research points are the control of the clutch and the control of the whole vehicle. For the vibration suppression when the motor is cut in, the widely used method is to achieve the stability of mode switching and power compensation through the VCU coordinated control of HEV. For example: using the disturbance observer to estimate and compensate for the disturbance to improve the control accuracy, tracking performance and driving performance; adding a mean model to the vehicle controller to estimate the torque, controlling the motor to compensate for the overall torque error, and improving the driving experience during mode switching. However, for VCU coordinated control, the switching process often involves multiple controller unit communications based on CAN. However, the application of CAN will not only cause network communication delays, but also bring protocol constraints. Communication delays will reduce the performance of the control algorithm and even make the system unstable. Many scholars have done a lot of research work in this area, such as: proposing a design method that combines an adaptive sliding mode controller with a fuzzy-based dynamic sampling cycle scheduling strategy to change the characteristics of the communication process and then deal with the delay problem; using an unscented Kalman predictor to fuse prior information with sensor signals to form motion perception, thereby compensating for time lag and noise, etc. Although the delay problem has been improved, the computing burden of the VCU has greatly increased, and the delay problem of CAN communication still exists objectively. Summary of the invention

[0005] Technical problem: In view of the shortcomings of the existing technology, a method for suppressing the oscillation of the hybrid transmission mode switching by actively responding to the motor based on MPC is provided. By actively responding to the MPC algorithm in the oscillation suppression controller, the motor actively responds to the target speed, so that the clutch friction plate can quickly engage at the synchronous speed, and the engagement is simplified. After the engagement is completed, the motor quickly switches to torque control, and outputs the target torque based on the power demand given by the vehicle controller, thereby achieving rapid power output. On the basis of effectively suppressing the oscillation, the communication delay problem in the traditional mode switching method is reduced, making the power output more continuous.

[0006] Technical solution: To achieve the above technical objectives, the present invention provides an MPC-based motor active response hybrid transmission mode switching oscillation suppression method, which is used for fast following of motor speed and fast power compensation when the power mode of a hybrid vehicle is switched, including a pure internal combustion engine power output mode, a power output mode of the internal combustion engine and the main motor at the same time, and a power output mode of the internal combustion engine main motor and the second motor intervening at the same time; including a vehicle controller, a clutch controller, and an oscillation suppressor, wherein the vehicle controller controls the intervention of the second motor to the clutch controller through the oscillation suppressor, wherein the oscillation suppressor includes a prediction control module and a torque compensation module, and the prediction control module includes an M PC algorithm unit and judgment program unit; when the vehicle controller issues a power mode and power demand that requires the intervention of the second motor according to the driver's needs, the predictive control module and the torque compensation module in the oscillation suppressor cooperate with each other to finally output a reference voltage to the PWM wave modulator, so that the speed of the second motor is actively responded to the speed of the clutch active plate, so that the clutch engagement process is fast and smooth, that is, the system oscillation caused by the intervention of the second motor is reduced, and the torque compensation module is used to realize the rapid power compensation of the second motor, while avoiding the communication delay problem, greatly improving the intervention speed of the motor power, and realizing efficient and smooth power enhancement response of the whole vehicle.

[0007] Furthermore, when the predictive control module receives the operating mode switching command, the Hall sensor is used to detect the speed of the clutch active plate ω r , the clutch active disc speed ω r As the reference speed of the predictive control module, the continuous set model unit is then used to generate a reference voltage, and then the judgment program unit is entered to analyze, process and select the data. If there is a speed difference between the friction plates, it is judged that the second motor is mainly controlled by speed at this time, and the clutch friction plate has not yet reached the synchronous speed with the second motor. If the clutch and the second motor are engaged at this time, an impact will be generated, causing obvious vibrations. Therefore, the clutch controller controls the second motor and the clutch to be separated. If the speed difference is zero, it indicates that the clutch and the second motor have reached the synchronous speed, then the clutch controller controls the clutch and the second motor to engage. After the second motor is engaged with the clutch, the second motor is controlled to switch to the torque control mode. At this time, the voltage value locked by the predictive control module is output to the torque compensation module as the final output reference value to ensure smooth switching of the torque compensation module.

[0008] Furthermore, the continuum model includes a permanent magnet synchronous motor PMSM state space model, a multi-step prediction model, a cost function and constraints of a model predictive control MPC algorithm;

[0009] ①PMSM state space model:

[0010] In the continuous case, the state space continuous model of PMSM is:

[0011]

[0012] Where: e =pω m , is the electrical angular velocity; ω m is the mechanical angular velocity; i d and i q are the stator d-axis and q-axis currents respectively; u d and u q are the stator d-axis and q-axis voltages respectively; L d and L q are the stator d-axis and q-axis equivalent inductances respectively; R s is the stator resistance; p is the number of pole pairs; J is the motor moment of inertia; B is the friction coefficient; T l is the load torque; f is the permanent magnet flux of the motor;

[0013] There is still dynamic coupling between the dq axis current and the speed. After decoupling, there is a nonlinear coupling term, which needs to be further processed. d i q , using SPMSM, that is, the stator d-axis and q-axis equivalent inductance are equal L d =L q When d i q The term can be eliminated; for the quadratic term ω e i q and ω e i d , at the PMSM operating point [i q0 ω e0 i d0 ] to perform a first-order Taylor expansion of these two terms:

[0014]

[0015] Then the state space equation of PMSM is:

[0016]

[0017] Where x = [i d i q ω e ] T ,u=[u d u q ] T , y=[i d ω e ] T ; A c , Bc , C c is the coefficient matrix;

[0018]

[0019] The above continuous model is transformed into a discrete state space model using Euler forward discretization, with a sampling time of T s :

[0020]

[0021] Among them, A d =T s *A c +I, B d =T s *B c , C c =C d ,

[0022] Then, the interference and error term εT are eliminated by difference s ,make:

[0023]

[0024] Then transform formula (4) into:

[0025]

[0026] Create new state variables and input quantities:

[0027]

[0028] The new state space equation for the second motor is:

[0029]

[0030] in,

[0031]

[0032] ②Multi-step prediction model:

[0033] In formula (7), y(k) represents the output variable matrix at the current moment [i d ω e ] T , N after k time p The output quantities are:

[0034]

[0035] Among them, N p is the prediction step length, N c is the control step size, and Np >N c ,make:

[0036]

[0037] Then formula (8) can be simplified to

[0038] Y I =ΩX(k)+ΓU I (9)

[0039] in

[0040]

[0041] ③ Cost function

[0042] At the sampling time T of the continuous set model prediction s By solving the quadratic programming problem, the cost function is minimized. The multi-input multi-output system MIMO is used for quadratic programming and the following solution is adopted:

[0043] min(Z T VZ+W T Z+D) (10)

[0044] Among them, Z contains variables, and the other matrices are constant matrices. The first term is nonlinear, the middle term is linear, and the last term is a constant. If V is a quadratic matrix, the quadratic programming problem can be simplified to a least squares problem, that is, the cost function is simplified to a quadratic programming and then obtained;

[0045] For different situations, the MPC algorithm unit needs to have different processing methods, and the cost function of formula (9) is defined as:

[0046]

[0047] Where, Q = diag(Q1,Q2···Q Np ) is the weight coefficient of the input to the MPC controller, R = diag(R1, R2···R Nc ) is the weight coefficient of the MPC controller output, Y Ref is the reference quantity; since there are two quadratic terms in formula (11), formula (9) is brought into formula (11) for simplification:

[0048]

[0049] Where G = R + Γ T QΓ, F=Γ T Q, H = Y Ref —ΩX(k). In this way, equation (11) is simplified to the standard form of quadratic programming, and the quadratic programming problem is solved iteratively;

[0050] The result of quadratic programming is U *I , take the first set of optimal solutions as the final output variables, that is:

[0051]

[0052] According to formula (5), the optimal output at time k is:

[0053] U out (k) = U out (k-1)+ΔU out (k) (14)

[0054] ④ Constraints of the MPC algorithm:

[0055] The constraints of the MPC algorithm are the x and u vectors, u d and u q Within the voltage range of the converter, using the stator q-axis current i q and the stator d-axis current i d The prediction model, namely, equation (15), transforms the current constraint into the stator d-axis voltage u d and stator q-axis voltage u q Constraints: From equation (1) and Euler forward discretization, we know that the stator q-axis current i q and the stator d-axis current i d The predicted value at the next moment is:

[0056]

[0057] According to formula (15), i q (k+1) and i d The constraint of (k+1) is transformed into u d (k) and u q (k) constraints, namely:

[0058]

[0059] in, They are the maximum and minimum values ​​corresponding to ud and uq respectively. i q (k+1) and i d The constraint of (k+1) is transformed into u d 、u q The corresponding range;

[0060] Generate a reference voltage U through the MPC algorithm unit d1 , U q1 At the same time, the voltage value at this moment is locked through the delay link to obtain the transition variable U d1 , U q1 , Δω enters the judgment program unit for data analysis, processing and selection. If the speed difference exists, the motor control method should be based on speed control. The U output by the MPC algorithm d1 , U q1 is the voltage of the next stage, namely U d0 =U d1 , U q0 =U q1 ; At this time, the clutch friction plate has not yet reached the synchronous speed. If it is engaged, it will produce a great impact and cause obvious vibration. Therefore, the clutch control signal is disengaged. If the speed difference is zero, it means that the clutch friction plate has reached the synchronous speed. Then the clutch control signal is engaged, and the motor control quickly switches to the torque control mode after engagement. At this time, the task of the predictive control module is nearly completed. In order to make the switching process of the torque compensation module smooth, the voltage value locked at this moment needs to be The reference value as the final output is output to the torque compensation module.

[0061] Furthermore, the oscillation suppressor uses the power demand as the reference torque Tref of the torque compensation module. In the active response phase, the motor speed needs to be quickly increased to the clutch active plate speed ω r Therefore, the motor control is mainly based on speed control. At this stage, the clutch engagement state is in the disengaged state. The power demand Tref of the vehicle controller and the clutch engagement state are ANDed to obtain the transition variable Tref* of the reference torque. At this time, Tref*=0. The large torque current ratio MPTA equation is used to obtain the reference of the current loop in FOC control. The voltage secondary reference value U is obtained through the proportional integral controller PI. d2 , U q2 , and then U d2 , U q2 , clutch engagement state and output U d0 , U q0 The data is input into the judgment program unit for data analysis, processing and selection; at this stage, the clutch logic state is 0, the power demand is reset to zero after AND operation, and the output of the final torque compensation module is U d , U q The output of the predictive control module U d0 , U q0 equal.

[0062] Further, when the speed difference Δω is 0, the clutch and the second motor are quickly engaged. At this time, the output of the prediction control module is U d0 , U q0 is the MPC output value locked at Δω=0 In this way, a smooth transition between speed control and torque control is achieved; after the clutch is fully engaged with the second motor, the clutch active plate, the driven plate and the output shaft of the motor can be regarded as a rigid body, and the clutch logic state becomes 1; at this time, the reference torque transition variable Tref* obtained by AND operation of the power demand reference Tref and the clutch engagement state is equal to the reference torque Tref, and Tref*, the clutch engagement state and the output U of the predictive control module are combined. d0 , U q0 The clutch engagement state is 1. Tref* is passed through the Ramp module to prevent torque mutation. The output of Ramp is used as the current loop reference after passing through the MPTA equation. Finally, the PI controller and the current controller output the voltage secondary reference value U. d2 , U q2 At this time, the judgment program unit will output the torque compensation module U d , U q Assign U d2 , U q2 , full control of the second motor is achieved through the modulator.

[0063] Further, the derivation process of the MPTA equation is:

[0064] The electromagnetic torque equation of permanent magnet synchronous motor is:

[0065] T e =1.5pi q [Ψ f +i d (L d -L q )] (17)

[0066] Where: p is the pole pair number, Ψ f is the permanent magnet flux, L d , L q They are the d-axis inductance and the q-axis inductance respectively.

[0067] The current limit circle and equal current circle of the permanent magnet synchronous motor when it is running are:

[0068]

[0069]

[0070] Combining (17) and (19), we can get the expression of Te / is, and find its first-order partial derivative:

[0071]

[0072] Substituting equations (17), (18) and (19) into (20), we have:

[0073]

[0074] If the motor is a hidden pole type, Ld = Lq, id = 0, its MPTA trajectory is:

[0075]

[0076] Furthermore, the final output voltage U of the torque compensation module is d , U q Perform Clarke-Park coordinate inverse transformation to obtain the output value u of the three-phase voltage a 、u b 、u c , input the output value into SVPWM modulation, then output a specific vector to the converter, and perform a specific switching action on the switch tube to achieve FOC control of the motor;

[0077] For Clarke transformation, the α-axis and the a-axis are defined to be collinear, and the α-axis lags the β-axis by 90°. According to the principle of equal-amplitude transformation, the transformation matrix is:

[0078]

[0079] The inverse transformation matrix is:

[0080]

[0081] For Park transformation, the α-β coordinate system is transformed into the dq synchronous rotating coordinate system:

[0082]

[0083] The inverse transformation matrix is:

[0084]

[0085] Where θ is the electrical angle of the second motor at this time.

[0086] Beneficial effects: Compared with the prior art, the motor active response hybrid transmission mode switching oscillation suppression method proposed in the present invention based on model predictive control has the following advantages:

[0087] 1. Using advanced control algorithms, the motor response speed is greatly improved, making the meshing process faster;

[0088] 2. Through direct control of the motor controller MCU, the delay problem caused by CAN communication is fundamentally eliminated, so that the torque compensation after the motor is engaged is faster. At the same time, the Ramp module is used to make the power compensation process smoother;

[0089] 3. The present invention has high universality and is applicable to hybrid vehicles with different purposes and load scales and power systems of different configurations under similar working conditions.

[0090] This method sinks the speed demand calculation that was previously completed in the VCU to the MCU for implementation, greatly reducing the VCU's computing and control burden while reducing the number of communications between the VCU and the MCU, bypassing the impact of CAN communication delay on system response and stability, and achieving rapid response of the motor and rapid torque compensation after engagement. The use of model predictive control algorithms makes the motor's active response faster, the power output more continuous, and the driving and riding experience of passengers better. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 It is a control block diagram of the motor active response hybrid transmission mode switching oscillation suppression method based on MPC (Model Predictive Control) of the present invention;

[0092] Figure 2 is a schematic diagram of a predictive control module of the present invention;

[0093] Figure 3 It is the MPC principle diagram of the present invention;

[0094] Figure 4 It is the hybrid optimization principle diagram of the present invention;

[0095] Figure 5 is a schematic diagram of a torque compensation module of the present invention;

[0096] Figure 6 This is a schematic diagram of the closed-loop step response of the current loop without the Ramp module;

[0097] Figure 7 Schematic diagram of the closed-loop response curve of the current loop under different slopes of the present invention;

[0098] Figure 8 This is a schematic diagram comparing the control effects of MPC and PI;

[0099] Fig. 9 It is a schematic diagram of the switching curve of the MPC and torque compensation joint control mode; DETAILED DESCRIPTION

[0100] The present invention will be further described below in conjunction with the accompanying drawings:

[0101] The method for suppressing oscillation of motor active response hybrid transmission mode switching based on model predictive control includes the following steps:

[0102] 1) The vehicle controller issues a mode switching command and power demand;

[0103] 2) The prediction control module works and outputs a reference voltage;

[0104] 3) The torque compensation module outputs the final reference voltage;

[0105] 4) FOC control through modulation;

[0106] When the vehicle is in the state of internal combustion engine power output, if the load suddenly increases or the energy battery is fully charged, Figure 1 The VCU can issue a mode switching command to switch the vehicle from internal combustion engine mode to hybrid mode. Or in hybrid mode, if further power boost is needed, the VCU can also issue a command to enable the second motor to intervene. This command is divided into a mode switching command and a power demand T ref .

[0107] Figure 1 The predictive control module in the control module starts to work. The detailed control block diagram is as follows Figure 2 As shown, the Hall sensor can be used to detect the speed of the clutch active plate ω r , which is used as the reference speed of the predictive control module. This part is based on model predictive control. Figure 3 This is the basic principle diagram of MPC. At sampling time k, the outputs of the next P moments are predicted based on the reference value and the state value at this moment. These outputs are considered to be the optimal outputs for subsequent control from the k moment. P is called the prediction step. The optimal output is only implemented for the first time at time k+1. After the execution is completed, the system enters the prediction work at time k+1. The steps are the same as at time k. In this way, the results and outputs of the previous shot are called the prediction inputs of this shot, thus realizing "rolling optimization". Figure 4 shown.

[0108] The system of the present invention is modeled below.

[0109] ① Establish PMSM state space model

[0110] In the continuous case, the state space equation of the PMSM is:

[0111]

[0112] Where: e =pω m , is the electrical angular velocity; ω m is the mechanical angular velocity; i d and i qare the stator d-axis and q-axis currents respectively; u d and u q are the stator d-axis and q-axis voltages respectively; L d and L q are the stator d-axis and q-axis equivalent inductances respectively; R s is the stator resistance; p is the number of pole pairs; J is the motor moment of inertia; B is the friction coefficient; T l is the load torque; f is the permanent magnet flux of the motor.

[0113] There is still dynamic coupling between the dq axis current and the speed. After decoupling, there are nonlinear coupling terms, which need to be further processed. d i q , this paper adopts SPMSM, namely L d =L q ,i d i q The term can be eliminated. e i q and ω e i d , at the PMSM operating point [i q0 ω e0 i d0 ] to perform a first-order Taylor expansion of these two terms:

[0114]

[0115] The state space equation of PMSM can be written as:

[0116]

[0117] Where x = [i d i q ω e ] T ,u=[u d u q ] T , y=[i d ω e ] T ;

[0118]

[0119] The above continuous model is transformed into a discrete state space model using Euler forward discretization, with a sampling time of T s :

[0120]

[0121] Among them, Ad =T s *A c +I, B d =T s *B c , C c =C d For the interference and error term εT that still exist s , which can be eliminated by making a difference.

[0122] make:

[0123]

[0124] Then formula (4) can be transformed into:

[0125]

[0126] Create new state variables and input quantities:

[0127]

[0128] The new state space equation is:

[0129]

[0130] in,

[0131]

[0132] ②Multi-step prediction model

[0133] In formula (7), y(k) represents the output variable at the current moment, and N after moment k p The output quantities are:

[0134]

[0135] Among them, N p is the prediction step length, N c is the control step size, and N p >N c ,make:

[0136]

[0137] Then formula (8) can be simplified to

[0138] Y I =ΩX(k)+ΓU I (9)

[0139] in

[0140]

[0141] ③Cost function and its optimization

[0142] At the sampling time T of the continuous set model prediction s The cost function is minimized by solving the quadratic programming problem. The system of the present invention is a multiple-input multiple-output (MIMO) system, and the quadratic programming problem of the MIMO system is generally solved in a standard form:

[0143] min(Z T VZ+W T Z+D)(10)

[0144] Among them, Z contains variables, and the rest of the matrices are constant matrices. The first term is nonlinear, the middle term is linear, and the last term is a constant. If V is a quadratic matrix, the quadratic programming problem is simplified to a least squares problem. In short, after simplifying the cost function to a quadratic programming, the problem can be solved by an existing program.

[0145] The cost function of this system is defined as:

[0146]

[0147] Where, Q = diag(Q1,Q2···Q Np ), R=diag(R1,R2···R Nc ), are the weight coefficients of input and output respectively, Y Ref is the reference quantity. Since there are two quadratic terms, it is necessary to further simplify it. Substituting equation (9) into equation (11), after simplification, we have:

[0148]

[0149] Where G = R + Γ T QΓ, F=Γ T Q, H = Y Ref —ΩX(k). Thus, equation (11) is simplified to the standard form of quadratic programming. Using the existing program, the quadratic programming problem can be solved iteratively.

[0150] The result of quadratic programming is U *I , take the first set of optimal solutions as the final output variables, that is:

[0151]

[0152] According to formula (5), the optimal output at time k is:

[0153] U out (k) = U out(k-1)+ΔU out (k) (14)

[0154] ④ Constraints

[0155] The constraints of MPC are the x and u vectors, u d and u q Within the voltage range of the converter; i q Mainly constrained by the maximum torque of the motor; i d The range of is determined by the MPTA equation or directly taken as 0. However, i q It is not possible to directly constrain this algorithm, and i q Too large a value may cause stator overcurrent or motor speed overshoot.

[0156] Use i q and i d The prediction model converts the current constraints into u d and u q Constraints. From formula (1) and Euler forward discretization, we can know that i q 、i d The predicted value at the next moment is:

[0157]

[0158] According to formula (15), i q (k+1) and i d The constraint of (k+1) is transformed into u d (k) and u q (k) constraints, namely:

[0159]

[0160] in, They are the maximum and minimum values ​​corresponding to ud and uq respectively. i q (k+1) and i d The constraint of (k+1) is transformed into u d 、u q The corresponding range.

[0161] Through the MPC algorithm, a reference voltage U is generated d1 , U q1 At the same time, the value at this moment is locked through the delay link to obtain U d1 , U q1, Δω enters the judgment program unit for data analysis, processing and selection. If Δω≠0, the motor control method should be mainly speed control. The U output by the MPC algorithm d1 , U q1 is the voltage of the next stage, namely U d0 =U d1 , U q0 =U q1 At this time, the clutch friction plate has not yet reached the synchronous speed. If it is engaged, it will produce a great impact and cause obvious vibration. Therefore, the clutch control signal is disengagement clutch*=0.

[0162] If Δω=0, it means the clutch friction plate has reached synchronous speed, then the clutch control signal is engaged clutch*=1, and the motor control quickly switches to torque control mode after engagement. At this time, the task of the predictive control module is nearly completed. In order to make the switching process of the next module smooth, the voltage value locked at this moment needs to be The reference value as the final output is output to the torque compensation module.

[0163] The power demand is used as the reference torque Tref of the torque compensation module. However, in the active response phase, the motor speed needs to be quickly increased to the clutch active plate speed ω r Therefore, the motor control is mainly based on speed control. At this stage, the clutch engagement state is clutch = 0, and the power demand Tref of the vehicle controller and the clutch engagement state are ANDed to obtain Tref*, such as Figure 5 As shown, at this time Tref*=0. Using the MPTA equation, the reference of the current loop is obtained. The voltage secondary reference value U is obtained through the PI controller d2 , U q2 . Then U d2 , U q2 , clutch and output U d0 , U q0 The data is input into the judgment program unit for data analysis, processing and selection. In this stage, clutch = 0, the power demand is reset to zero after AND operation, and the output of the final torque compensation module is U d , U q The output of the predictive control module U d0 , U q0 equal.

[0164] When the speed difference Δω is zero, clutch*=1, and the clutch engages quickly. At this time, the output of the prediction control module is U d0 , U q0 is the MPC output value locked at Δω=0 This achieves a smooth transition between speed control and torque control. After the clutch is fully engaged, the clutch active plate, driven plate and the output shaft of the motor can be regarded as a rigid body, clutch = 1. At this time, the power demand reference Tref and the clutch engagement state are ANDed to obtain Tref*, which is equal to Tref. d0 , U q0 Input them into the judgment program unit together, at this time, clutch = 1. Pass Tref* through the Ramp module to prevent torque mutation, and use the output of Ramp as the current loop reference after passing through the MPTA equation, and finally pass through the PI controller to output the secondary voltage reference value U d2 , U q2 At this time, the judgment program unit will output the torque compensation module U d , U q Assign U d2 , U q2 . Full control of the motor is achieved through the modulator.

[0165] Derivation of the MPTA equation

[0166] The electromagnetic torque equation of permanent magnet synchronous motor is:

[0167] T e =1.5pi q [Ψ f +i d (L d -L q )] (17)

[0168] Where: p is the pole pair number, Ψ f is the permanent magnet flux, L d , L q They are the d-axis inductance and the q-axis inductance respectively.

[0169] The current limit circle and equal current circle of the permanent magnet synchronous motor when it is running are:

[0170]

[0171]

[0172] Combining (17) and (19), we can get the expression of Te / is, and find its first-order partial derivative:

[0173]

[0174] Substituting equations (17), (18) and (19) into (20), we have:

[0175]

[0176] If the motor is a hidden pole type, Ld = Lq, id = 0, its MPTA trajectory is:

[0177]

[0178] In this module, the parameter that has the greatest impact on the system's rapid response and stability is the slope k of the Ramp module. Figure 6 This is the closed-loop response curve of the system when there is no Ramp module. It can be seen that although the system tends to be stable in the end, the overshoot is too large and the oscillation time is too long, and the subsequent rapid power response cannot be achieved. Therefore, a Ramp module needs to be added to buffer the input. The value of k determines the quality of the buffering effect and the length of the stabilization time. In order to study the impact of k on the response speed and stability of the system, k = [1, 3, 5, 10, 15, 20] is taken for analysis. The current loop closed-loop response curves under different values ​​are shown in the figure below. Figure 7 As shown. It can be seen that if the value of k is too small, the response time will be too long and the response speed of the system will be slow; if the value of k is too large, the overshoot of the system will increase, affecting the stability of the system. Therefore, in the system adopted by the present invention, it is better for k to be between [3,10].

[0179] Final output voltage U d , U q Perform Clarke-Park coordinate inverse transformation to obtain the output value u of the three-phase voltage a 、u b 、u c , enters SVPWM modulation, and finally outputs a specific vector to the converter, performs a specific switching action on the switch tube, thereby realizing FOC control of the motor. The coordinate transformation formula is as follows.

[0180] For Clarke transformation, the α-axis and the a-axis are defined to be collinear, and the α-axis lags the β-axis by 90°. According to the principle of equal-amplitude transformation, the transformation matrix is:

[0181]

[0182] The inverse transformation matrix is:

[0183]

[0184] For Park transformation, the α-β coordinate system is transformed into the dq synchronous rotating coordinate system:

[0185]

[0186] The inverse transformation matrix is:

[0187]

[0188] The proposed method is modeled and simulated using MATLAB / simulink. The motor speed waveform under the MPC control strategy is shown in the figure below. Figure 8 As shown in the figure, it can be seen that for the same motor control object, the overshoot of the response curve under MPC is almost zero, the rise time is close, and the stabilization time is much shorter than the effect under PI control. Therefore, the advantage of the MPC algorithm is very obvious. Fig. 9 The figure shows the torque and speed waveforms of the motor when the mode is switched under the combination of MPC control and torque compensation. By adding torque control on the basis of MPC's fast response, the speed waveform is stable, the oscillation during the meshing process is almost zero, and the torque compensation is rapid, which verifies the feasibility of the oscillation suppression strategy based on MPC.

Claims

1. A method for suppressing oscillations of hybrid transmission mode switching by active motor response based on MPC, Features: Used for fast following of motor speed and fast power compensation when power mode switching of hybrid electric vehicles, including pure internal combustion engine power output mode, internal combustion engine and main motor output power mode at the same time, and internal combustion engine main motor and second motor intervention power mode at the same time; including vehicle controller, clutch controller, and oscillation suppressor, wherein the vehicle controller controls the intervention of the second motor to the clutch controller through the oscillation suppressor, wherein the oscillation suppressor includes a prediction control module and a torque compensation module, and the prediction control module includes an MPC algorithm unit and a judgment program unit; when the vehicle controller issues a power mode and power demand requiring the intervention of the second motor according to the needs of the driver, the prediction control module and the torque compensation module in the oscillation suppressor cooperate with each other to finally output a reference voltage to the PWM wave modulator, so that the speed of the second motor actively responds to the speed of the clutch active plate, so that the clutch engagement process is fast and smooth, that is, the system oscillation caused by the intervention of the second motor is reduced, and the torque compensation module is used to realize the fast power compensation of the second motor, while avoiding the communication delay problem, greatly improving the intervention speed of the motor power, and realizing the efficient and smooth response of the vehicle power enhancement; When the predictive control module receives the operating mode switching command, the Hall sensor is used to detect the speed of the clutch active plate ω r , the clutch active disc speed ω r As the reference speed of the prediction control module, the continuous set model unit is then used to generate a reference voltage, and then the judgment program unit is entered to analyze, process and select the data. If there is a speed difference between the friction plates, it is judged that the second motor is mainly controlled by speed control at this time, and the clutch friction plate has not yet reached the synchronous speed with the second motor. If the clutch and the second motor are engaged at this time, an impact will be generated, causing obvious vibrations. Therefore, the clutch controller controls the second motor and the clutch to be separated. If the speed difference is zero, it indicates that the clutch and the second motor have reached the synchronous speed, then the clutch controller controls the clutch and the second motor to engage. After the second motor is engaged with the clutch, the second motor is controlled to switch to the torque control mode. At this time, the voltage value locked by the prediction control module is output to the torque compensation module as the final output reference value to ensure that the switching process of the torque compensation module is smooth; The continuum model includes a permanent magnet synchronous motor PMSM state space model, a multi-step prediction model, a cost function and constraints of a model predictive control MPC algorithm; ①PMSM state space model: In the continuous case, the state space continuous model of PMSM is: Where: e =pω m , is the electrical angular velocity; ω m is the mechanical angular velocity; i d and i q are the stator d-axis and q-axis currents respectively; u d and u q are the stator d-axis and q-axis voltages respectively; L d and L q are the stator d-axis and q-axis equivalent inductances respectively; R s is the stator resistance; p is the number of pole pairs; J is the motor moment of inertia; B is the friction coefficient; T l is the load torque; f is the permanent magnet flux of the motor; There is still dynamic coupling between the dq axis current and the speed. After decoupling, there is a nonlinear coupling term, which needs to be further processed. d i q , using SPMSM, that is, the stator d-axis and q-axis equivalent inductance are equal L d =L q When d i q The term can be eliminated; for the quadratic term ω e i q and ω e i d , at the PMSM operating point [i q0 ω e0 i d0 ] to perform a first-order Taylor expansion of these two terms: Then the state space equation of PMSM is: Among them, x=[i d i q ω e ] T , u = [u d u q ] T , y=[i d ω e ] T ;A c , B c , C c Main system number square; The above continuous model is transformed into a discrete state space model using Euler forward discretization, with a sampling time of T s : wherein, A d = T s * A c + I, B d = T s * B c , C d = C c , Then, the interference and error term εT are eliminated by difference s ,make: Then transform formula (4) into: Create new state variables and input quantities: The new state space equation for the second motor is: in, ②Multi-step prediction model: In formula (7), y(k) represents the output variable matrix at the current moment [i d ω e ] T , N after k time p The output quantities are: Among them, N p is the prediction step length, N c is the control step size, and N p >N c ,make: Then formula (8) can be simplified to AND I =ΩX(k)+ΓU I (9) in ③ Cost function At the sampling time T of the continuum model prediction s By solving the quadratic programming problem, the cost function is minimized. The multi-input multi-output system MIMO is used for quadratic programming and the following solution is adopted: min(Z T VZ+W T Z+D)(10) Among them, Z contains variables, and the other matrices are constant matrices. The first term is nonlinear, the middle term is linear, and the last term is a constant. If V is a quadratic matrix, the quadratic programming problem can be simplified to a least squares problem, that is, the cost function is simplified to a quadratic programming and then obtained; For different situations, the MPC algorithm unit needs to have different processing methods, and the cost function of formula (9) is defined as: Where, Q = diag (Q 1 ,Q 2 ···Q Np ) is the weight coefficient of the input to the MPC controller, R = diag(R 1 ,R 2 ···R Nc ) is the weight coefficient of the MPC controller output, Y Ref is the reference quantity; since there are two quadratic terms in formula (11), formula (9) is brought into formula (11) for simplification: Where G = R + Γ T QΓ,F=Γ T Q, H = Y Ref -ΩX(k); Thus, formula (11) is simplified to the standard form of quadratic programming, and the quadratic programming problem is solved iteratively; The results of quadratic programming Take the first set of optimal solutions as the final output variables, that is: According to formula (5), the optimal output at time k is: AT out (k)=U out (k-1)+ΔU out (k) (14) ④ Constraints of the MPC algorithm: The constraints of the MPC algorithm are the x and u vectors, u d and u q Within the voltage range of the converter, using the stator q-axis current i q and the stator d-axis current i d The prediction model, namely, equation (15), transforms the current constraint into the stator d-axis voltage u d and stator q-axis voltage u q Constraints: From equation (1) and Euler forward discretization, we know that the stator q-axis current i q and the stator d-axis current i d The predicted value at the next moment is: According to formula (15), i q (k+1) and i d The constraint of (k+1) is transformed into u d (k) and u q (k) constraints, namely: in, u d 、u q The corresponding maximum and minimum values ​​are i q (k+1) and i d The constraint of (k+1) is transformed into u d 、u q The corresponding range; Generate a reference voltage U through the MPC algorithm unit d1 , U q1 At the same time, the voltage value at this moment is locked through the delay link to obtain the transition variable U d1 , U q1 , The speed difference Δω enters the judgment program unit for data analysis, processing and selection. If the speed difference exists, the motor control method should be based on speed control. The U output by the MPC algorithm d1 , U q1 is the voltage of the next stage, namely U d0 =U d1 , U q0 =U q1 ; At this time, the clutch friction plate has not reached the synchronous speed. If it is engaged, it will produce a great impact and cause obvious vibration. Therefore, the clutch control signal is disengaged. If the speed difference is zero, it means that the clutch friction plate has reached the synchronous speed. Then the clutch control signal is engaged, and the motor control quickly switches to the torque control mode after engagement. At this time, the task of the predictive control module is nearly completed. In order to make the switching process of the torque compensation module smooth, the voltage value locked at this moment needs to be The reference value as the final output is output to the torque compensation module.

2. According to the MPC-based motor active response hybrid transmission mode switching oscillation suppression method of claim 1, Features: The oscillation suppressor uses the power demand as the reference torque Tref of the torque compensation module. In the active response phase, the motor speed needs to be quickly increased to the clutch active plate speed ω r Therefore, the motor control is mainly based on speed control. At this stage, the clutch engagement state is in the disengaged state. The power demand Tref of the vehicle controller and the clutch engagement state are ANDed to obtain the transition variable Tref* of the reference torque. At this time, Tref*=0. The large torque current ratio MPTA equation is used to obtain the reference of the current loop in FOC control. The voltage secondary reference value U is obtained through the proportional integral controller PI. d2 , U q2 , and then U d2 , U q2 , clutch engagement state and output U d0 , U q0 The data is input into the judgment program unit for data analysis, processing and selection; at this stage, the clutch logic state is 0, the power demand is reset to zero after AND operation, and the output of the final torque compensation module is U d , U q The output of the predictive control module U d0 , U q0 equal.

3. According to the MPC-based motor active response hybrid transmission mode switching oscillation suppression method of claim 2, Features: When the speed difference Δω is 0, the clutch and the second motor are quickly engaged. At this time, the output of the prediction control module is U d0 , U q0 is the MPC output value locked at Δω=0 In this way, a smooth transition between speed control and torque control is achieved; after the clutch is fully engaged with the second motor, the clutch active plate, the driven plate and the output shaft of the motor can be regarded as a rigid body, and the clutch logic state becomes 1; at this time, the reference torque transition variable Tref* obtained by AND operation of the power demand reference Tref and the clutch engagement state is equal to the reference torque Tref, and Tref*, the clutch engagement state and the output U of the predictive control module are combined. d0 , U q0 The clutch engagement state is 1. Tref* is passed through the Ramp module to prevent torque mutation. The output of Ramp is used as the current loop reference after passing through the MPTA equation. Finally, the PI controller and the current controller output the voltage secondary reference value U. d2 , U q2 At this time, the judgment program unit will output the torque compensation module U d , U q Assign U d2 , U q2 , full control of the second motor is achieved through the modulator.

4. According to claim 3, the motor active response hybrid transmission mode switching oscillation suppression method based on MPC, Features: The derivation process of the MPTA equation is: The electromagnetic torque equation of permanent magnet synchronous motor is: T e =1.5pi q [Ψ f +i d (L d -L q )] (17) Where: p is the pole pair number, Ψ f is the permanent magnet flux, L d , L q They are d-axis inductance and q-axis inductance respectively; The current limit circle and equal current circle of the permanent magnet synchronous motor when it is running are: i d 2 +i q 2 ≤I max 2 (18) Combining (17) and (19), we can get T e / i s The expression of , find its first-order partial derivative: Substituting equations (17), (18) and (19) into (20), we have: If the motor is a hidden pole type, Ld = Lq, id = 0, its MPTA trajectory is:

5. According to the MPC-based motor active response hybrid transmission mode switching oscillation suppression method of claim 4, Features: The final output voltage U d , U q Perform Clarke-Park coordinate inverse transformation to obtain the output value u of the three-phase voltage a 、u b 、u c , input the output value into SVPWM modulation, then output a specific vector to the converter, and perform a specific switching action on the switch tube to achieve FOC control of the motor; For Clarke transformation, the α-axis and the a-axis are defined to be collinear, and the α-axis lags the β-axis by 90°. According to the principle of equal-amplitude transformation, the transformation matrix is: The inverse transformation matrix is: For Park transformation, the α-β coordinate system is transformed into the dq synchronous rotating coordinate system: The inverse transformation matrix is: Where θ is the electrical angle of the second motor at this time.

Images