• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    MPC-based torque control of permanent magnet synchronous motor for electric vehicles via switching optimization

    2017-12-26 09:28:42BingtaoRENHongCHENHaiyanZHAOWeiXU
    Control Theory and Technology 2017年2期

    Bingtao REN,Hong CHEN?,Haiyan ZHAO,Wei XU

    1.State Key Laboratory of Automotive Simulation and Control,Jilin University,Changchun Jilin 130025,China;

    2.Department of Control Science and Engineering,Jilin University,Changchun Jilin 130025,China

    MPC-based torque control of permanent magnet synchronous motor for electric vehicles via switching optimization

    Bingtao REN1,2,Hong CHEN1,2?,Haiyan ZHAO1,2,Wei XU2

    1.State Key Laboratory of Automotive Simulation and Control,Jilin University,Changchun Jilin 130025,China;

    2.Department of Control Science and Engineering,Jilin University,Changchun Jilin 130025,China

    In order to effectively achieve torque demand in electric vehicles(EVs),this paper presents a torque control strategy based on model predictive control(MPC)for permanent magnet synchronous motor(PMSM)driven by a two-level three-phase inverter.A centralized control strategy is established in the MPC framework to track the torque demand and reduce energy loss,by directly optimizing the switch states of inverter.To fast determine the optimal control sequence in predictive process,a searching tree is built to look for optimal inputs by dynamic programming(DP)algorithm on the basis of the principle of optimality.Then we design a pruning method to check the candidate inputs that can enter the next predictive loop in order to decrease the computational burden of evaluation of input sequences.Finally,the simulation results on different conditions indicate that the proposed strategy can achieve a tradeoff between control performance and computational efficiency.

    Permanent magnet synchronous motor,electric vehicle,torque optimal control,model predictive control

    1 Introduction

    As an effective means to reduce pollution,electric vehicles(EVs)have drawn many attentions in the past few decades[1].Electric machines are the key component of EV technologies,which have to offer high efficiency,high controllability and wide-speed range.Due to high power density, high reliability and relatively smooth running,permanent magnet synchronous motor(PMSM)is popularly used for EVs as the motor type and has been actively developed by many research institutes[2–4].In order to effectively achieve drive demand of vehicle control system,the PMSM torque control plays a very important role in the EV technologies.

    To ensure fast and smooth torque response and high dynamic performance,many PMSM control methods have been studied in literatures.Field orientation control(FOC)and direct torque control(DTC)are two kinds of classical control schemes well used in alternating current(AC)motor system[5].It is difficult to handle multiple requirements and existing current constraints.In recent years,a novel control scheme via switching optimization has been studied[6–8],whose common idea is to directly manipulate the switch signals of inverter to control current or torque based on the predictive information of the system’s future behavior.The modulator is not required and the control pulse for inverter is a variable frequency signal.As an effective way to deal with a constrained optimal control problem,model predictive control(MPC)is chosen via switching optimization to satisfy diverse and even conflicting requirements in PMSM control system[9–11].It not only can produce instantaneous currents with high accuracy in a transient interval,but also reduce the switching losses by switching optimization which represent a major part of the energy losses of overall drive system.

    Due to the discrete switch states of modern power inverter,this MPC-based switching optimization scheme has to select the control input in a finite set of discrete values[6,12].By investigating all candidate sequences,the full enumeration method is a simple method to fast find the optimal switch sequence[13–15].But this method is only available for the MPC optimization scheme with one or two predictive horizons.The long predictive horizon greatly boosts the performance of MPC optimization,in the sense that the switching losses and the current or torque tracking error are further lower than that with short horizon.Meanwhile,as the increasing of predictive horizon,the number of candidate sequences may become very large and the computational burden will be heavy,full enumeration is not suitable for the case with long horizon. This switching combinational optimization in PMSM control system is a challenging problem that has led many researchers to study some efficient solution approaches.Reference[16]presents a control Lyapunov function to dynamically prune suboptimal sequences of search space to decrease the computational burden of MPC scheme.In[17],a predictive control technique with the predefined constraints is proposed to reduce the number of the available switch states of the voltage source inverter.Reference[18]adopts a branch and bound technique to achieve computationally efficient predictive direct torque control.These methods only explore some pruning methods and have not made a definite searching process to reduce some unnecessary computation and achieve a tradeoff between control performance and computation speed.

    To achieve fast and smooth torque demand of vehicle,this paper investigates a torque control strategy via switching optimization for PMSM system based on model predictive control.With respect to other methods considered in this presentation,the main merits of the proposed scheme cover the following points:1)The control scheme regards the motor and inverter as a complete controlled system,and directly optimizes the switch states of inverter.2)The MPC-based control strategy can simultaneously meet multiple requirements of EV propulsion,i.e.,to achieve fast and smooth torque demand,to reduce energy consumption and to deal with current saturation constraint.This leads to energy saving in a wide operating range and efficiency improvement while benefiting from the switching optimization advantages.3)Considering that dynamic programming(DP)[19]is avaiable for the solution to discrete-time sequential decision process,we use it to build a searching tree to look for optimal inputs in each stage on the basis of optimality principle.Then,to decrease the computational burden of evaluation of input sequences,a pruning method is designed to check candidate branches of this searching tree whether to enter the next predictive loop.The results in simulation demonstrate that it can effectively achieve torque control.

    The organization of this paper is as follows.In Section 2 a control oriented nonlinear model of the inverter and PMSM system is built.Section 3 introduces an MPC approach based on dynamic programming with branch and bound for motor torque control.Section 4 gives the simulation results to validate the effectiveness of the proposed strategy for different cases.Finally the conclusion and future work are summarized in Section 5.

    2 Modelling for control design

    The PMSM is driven by a two-level three-phase voltage source inverter from DC voltage source.They can be regarded as a complete controlled system,whose structure diagram is shown in Fig.1.To implement the torque control for PMSM system,an adequate system model is essential.

    Fig.1 Structure diagram of the PMSM system.

    2.1 Three phase inverter

    The three-phase inverter connects the three phases of motor to the DC positive or negative end and is composed by six power devices,like insulated-gate bipolar transistors(IGBT)that combines high efficiency and fast switching[20].

    The model of power device is considered as an ideal switch with only two states:on and off in[6].In order to avoid short circuiting connections,two switches in each inverter bridge in Fig.1 cannot operate at the same time.We define switching signal for each inverter bridge as Sa,b,cwith only two states 1 and 0:

    Sband Schave the same definition,and S=[SaSbSc]Tis defined as switching vector of inverter.Therefore,the total number of switch states of the inverter is equal to the number of different combinations of the two switch states of each switch,which has eight possible switch states with S∈{S0,S1,...,S7}described as Fig.2.The output voltage in each bridge is defined as

    where Vdcdenotes DC power supply voltage.

    Considering an unitary vectorwhich represents the phase displacement between two phases,to preserve the amplitude of phase voltage unchanged,the output voltage vector Vscan be defined as

    By substituting vector a into(3),the output voltage is expressed as a matrix form[VαVβ]Tin the α-β axis:

    Fig.2 Switch states.

    Frequent switching action can also cause some loss,hence inverter loss needs to be modeled here, consisting of switching loss and conduction loss.

    IGBT switching loss(turn-on and turn-off):

    IGBT conduction loss:

    Energy loss of one phase IGBT-bridge during one sample time:

    where EGon,EGoあdenote switching energy loss of turnon and turn-off of IGBT,Vnm,Inmare nominal voltage and current,and Vcedenotes gate threshold voltage.

    2.2 PMSM motor

    By the Park transformation,the stator voltage Vr=Vd+jVqin the rotor coordinate can be derived from the stator coordinate with Vr=Vse?jθe,which is expressed as a matrix form:

    where θeis electrical angle of the rotor.

    Based on Kirchoff’s voltage law and Newton’s second law,the current dynamics can be derived from equation(9)in the rotor direct and quadrature(d-q)coordinate.

    where Idand Iqdenote d and q axis stator currents respectively,Vdand Vqdenote d and q axis stator voltages,Ldand Lqdenote d and q axis inductances,ωrdenotes rotor shaft mechanical speed,Rmdenotes resistance of the stator windings,pmdenotes number of magnetic pole pairs of motor,φmdenotes the amplitude of the flux induced by the permanent magnets of the rotor in the stator phases.

    The electromagnetic torque Tein the machine depends on the flux magnitude φmand the stator current vectors Idand Iq,as follows:

    2.3 State space equation of the complete system

    We choose stator currents as system states with x=[IdIq]T,rotor speed as time varying parameter with d=ωrand switch inputs as control inputs with u=[SaSbSc]T.Hence,this is a special hybrid system with continuous states and discrete control inputs.Based on the related equations above(2),(4),(8)and(9),the space state model of the complete PMSM system can be developed:

    where Tsis afixed step size,is a sample time and t denotes running time.

    For performing a torque control design in real-time system,the system equation must be rewritten as a discrete-time equation used for calculation of the predictive output.Based on the forward Euler method,a discrete-time system model is obtained as

    3 Torque control design based on model predictive control

    In this section,we will design an MPC strategy via switching optimization to realize PMSM torque control.To look for optimal inputs, DP algorithm builds a searching process based on the optimality principle.Then a pruning branch method is designed to check the system stability of candidate inputs to reduce computational burden.

    3.1 Torque predictive control scheme

    The torque control scheme for PMSM via switching optimization is designed as shown in Fig.3,where the torque predictive controller can directly control the inverter and need no modulator.Hence the switch signals generated will have variable frequency based on current states,which may decrease the switching losses.Due to the finite number of control actions that are allowed at each time step,the problem of PMSM torque control can be defined as a combinatorial optimization that determines the next inverter switch state on the cost evaluation over a receding time horizon.

    Fig.3 MPC-based torque control structure.

    According to the MPC’s principle in[21,22],p is defined as predictive horizon to extend the forecast of the system future output.To simply the calculation,we set control horizon with the same value of predictive horizon p.At sample time k,U(k)is defined as the future control input sequence which is the optimization vector in the following cost function and X(k)is defined as the predictive state sequence,as follows:

    Each vector of U(k)and X(k)contains respectively an array of control inputs u and system states x. The reference current R(k)keeps the same value r in each predictive horizon.

    3.2 Design of the optimization function

    The main control requirements of PMSM are summarized as:

    .Tracking torque demand;and.Energy consumption reduction.

    Additionlly,current saturation constraints should be taken into account.

    MPC can effectively deal with an optimization problem with multi-objective,multi-constraint and multivari-able for the complex engineering system optimization design.These objectives can be discussed in details in the following part.

    The main control requirement is to ensure the PMSM tracking the torque demand from upper vehicle control system well.By using a rotating d-q reference frame oriented to the rotor magnetic field axis,each stator current component has a physical meaning.The real component Idis proportional to the reactive power,and the imaginary component Iqis proportional to the electromagnetic torque.

    In order to achieve an expected torque commandfrom upper controller,the q axis reference currentis calculated by equation(10).Meanwhile,to reduce the reactive power and the magnetic resistance torque,the d axis current should be small and even close to zero.The detailed equation is given in(13):

    Therefore,the torque control problem is converted as a current tracking control problem,which can also reduce the reactive power and the magnetic resistance torque.The cost function of this control problem is designed as

    Aiming at achieving good dynamic performance,the machine loss of PMSM system should be minimized in the cost function,which mainly consists of copper loss and inverter loss.The copper loss Emis produced by electrical currents in the conductors of transformer windings,which can be regarded as the heat energy of motor resistance[23].Additionally,frequent switching action of inverter can cause some power loss and switching oscillation,so energy losses of the inverter EGLneeds to be optimized,consisting of switching loss and conduction loss in equation(7),which can also reduce the number of switching action,thus indirectly avoiding the switch oscillation.Hence,the second cost function is designed to minimize energy losses of motor system during the total horizon:

    In addition,in order to guarantee the terminal state of the closed-loop system into the stability region,we consider a terminal state function in the cost function,which penalizes the state at the end of the predictive horizon.This quadratic terminal cost is formed as

    where P is a positive define and symmetric terminal penalty matrix.It should be pointed out that terminal penalty matrix P can be determined offline as[21],which can save the computing time of controller.

    As equations(14),(15)and(16)mentioned above,a total objective function of the optimal control problem with constraints is defined over a receding horizon,as follows:

    where Q and R denote weight matrixes of cost function.System dynamic constraint is to calculate the predictive state in each predictive horizon.The saturation constraint is an inequality constraint for each predictive state.

    3.3 Optimization solution based on DP and pruning method

    At the time instant k,if each predictive horizon is defined as each stage in control procedure,it is easy to observe from equation(17)that the MPC-based optimization is a multi-stage optimization problem.The optimal solution in Jkminimizes additive cost of each stage and satisfies the constraints of the states and inputs.Dynamic programming(DP)[19]is available for the solution to this discrete-time sequential decision process where the number of states is limited.Therefore,in order to fast solve this switching optimization problem in the MPC frame,we introduce the DP algorithm to build a searching tree to look for optimal input in each stage on the basis of optimality principle.

    According to the DP algorithm,the multi-stage problem of MPC(17)can be disassembled into a set of lower-dimensional problems.Since control input u has finite candidate solution in discrete set{S0,S1,...,S7}and has to choose the best from this finite number of possibility,this can be seen as a discrete combinational optimization problem.When a candidate control input is chosen at a given state,the next state is fully determined;that is,for any state x(k),control u(k),and time k,the next state x(k+1)is determined.

    In order to clearly analyze the optimization process,a searching tree is built for this multi-stage optimization problem.At simple time k,the current state is defined as the root of this searching tree.The candidate control input uj(k)is defined as one tree’s branch,where j=0,1,2,...,7.For different branches,the root grows different node xj(k+1)which is the predictive state for uj(k)in next stage.Every node contains the information of uj(k)and xj(k+1).While going to the next stage,the node also works as a father node to produce new nodes by choosing one branch from control candidate set.With the increasing of stages,the root x(k)will grow a large searching tree for the solution to optimization process.

    According to the principle of optimality,DP algorithm is a kind of depth-first traversal for optimization solution.For initial root x(k),the optimal cost is given by the last step of the following algorithm,which proceeds backward in time from stage p to stage 1.We re-define

    At stage m(1,2,...,p),the optimal cost is derived from

    where the derivation provides an interpretation of Jk+mas the optimal cost function at stage m.Subproblems obtained after decomposition are not independent of each other.The sub-problem in the previous stage is further solved on the basis of a solution on the next stage.To achieve optimal control sequence U?(k+m ? 1)in cost function Jk+m,it needs to select the candidate branch u(k+m)in discrete finite control set{S0,S1,...,S7}to minimize the cost function Jk+m+1,then addof suboptimal control signal u(k+m)to Jk+m,compare Jk+mof each candidate branch to get optimal u?(k+m ? 1),and therefore,the optimal vector U?(k+m ? 1)=[u?(k+m ? 1);U?(k+m)].

    This solution algorithm mainly consists of two parts,like predicting output process and looking for optimal input.More specifically,the framework of the solution algorithm of the switching optimization based on DP is:

    1)Initialize weight matrices Q,R,P,and stage p.

    2)Measure the current state x(k),i.e.,Id(k),Iq(k),and ωr(k).

    3)Set stage m=1.

    4)Choose candidate branch ujin order from the discrete control set{S0,S1,...,S7},where j=1,2,...,7,then calculate predictive node xj(k+m)by equation(12),finally m=m+1.

    5)If m<p,go back to step 4);else,go to next step.

    6)Calculate the function L(x,u)for each candidate branch uj(k+m?1)in order by equation(18),where j=0,1,...,7,then calculate the cost function Jk+mon the basis of L(x,u)and previous(if m=p,=0).Minimize Jk+mto select the optimal input u?(k+m ? 1)and determine suboptimal control signal sequence U?(k+m?1).Finally,m=m?1.

    7)If m≥1,go back to step 6);else,go to step 8).

    8)Apply the first optimal input u?(k)to the PMSM system.End this algorithm.

    Although DP algorithm can build a searching process of the real-time optimization solution,this method has to traverse each candidate signal in discrete control set,which is a heavy calculation burden.To reduce the number of switch states that enter the next predictive loop,we design a pruning method based on branch and bound strategy which is an algorithm design paradigm for discrete and combinatorial optimization problems[24].The algorithm explores the branches of this searching tree that represent subsets of the solution set by breadth-first searching.Therefore,a bounding function is designed to compute a lower bound for each node of the searching tree and woks as a problem-specific branching rule.While DP algorithm enumerates each candidate branch in one stage,the branch is checked by the bounding function,then if it cannot satisfy the branching rule,this branch is discarded and cannot run into next predictive stage.The searching tree can be shown as Fig.4.

    Fig.4 Pruning the searching tree.

    A Lyapunov function has been proposed in[16]to prune the candidate inputs,but the function is simple and loose to check the stability performance.We design a complicated Lyapunov function and give a clear calculation to check the stability performance of the closed loop system before using the candidate inputs to predict the future states.Define?x=x?r as the state of the closed-loop system.The Lyapunov function is defined as

    In order to ensure state?x=[0 0]Tof the closed-loop system is stable,the P0is positive definite matrix and the derivative along the trajectory of V(x)is set as˙V(?x)<0,as follows:

    Therefore,the bound function for pruning branch of searching tree in each horizon can be set as

    More specifically,the framework of this pruning branch algorithm in the predictive loop m(1,2,...,p)is:

    1)Initialize matrix P0and current state x(k+m?1).

    2)Choose one candidate input uj(k+m?1),j=0∶7.

    3)Calculate predictive state x(k+m)by equation(12)and state error?x(k+m)=x(k+m)?r(k).

    4)Calculate the derivative of the Lyapunov function,˙V(x(k+m+1))by equation(21).

    5)If˙V(x(k+m))<0 and I2q(k+m)+I2d(k+m)≤I2max,go to step 6);else,go to step 7).

    6)Come into next predictive loop m+1 and go back to step 2).

    7)Set the next cost functionof uj(k+m?1)as inf,stop going to next predictive loop,prune this candidate input in searching tree,and end this algorithm.

    At previous m stage,for each candidate control signal u,the algorithm checks the bound function in equation(22),if beyond the limit,this candidate control signal can be discarded,since it will not lead to the optimal solution;else,it is preserved and continues coming into next predictive loop.This would prune the number of inputs to enter the next predictive loop and decrease the computational burden of evaluation of input sequences.Hence,the judgement condition about the bound function is only used once time for each control inputs in each predictive stage.The flow diagram of the proposed whole algorithm is presented in Fig.5.The predictive horizon p=2 in flow diagram and u?(k+1)or u?(k+2)can be determined as the same process with the blue flow blocks of u?(k).Symbol j1,j2 denote the candidate solution numbers of the 1st and 2nd prediction stage,respectively.Itrequires multiple sets of tests toset these control parameters,like weight matrices Q and R.The optimal parameters are once determined and do not need to be updated online for different cases.Considering that both P0and Q affect tracking performance,we set P0is equal to matrix Q.Also,we calculate P for the terminal state function(16)as[21].

    Fig.5 Control flow diagram of the proposed algorithm.

    4 Simulation and analysis

    In this section,simulation results are shown to demonstrate the validity of the proposed MPC strategy.The simulation is evaluated on an PMSM offline simulation platform in Simulink/Matlab.The detailed parameters of this PMSM system are shown in Table 1.

    The reference torque input is set as a sinusoidal wave,in which the amplitude is 0.178N·m and the frequency is 1.59Hz.The rotor shaft mechanical speed is around 1200rad·s?1.The simulation time is set as 3s,the fixed step size Ts=0.1ms and predictive horizon p=3.The responses are compared between the cases before pruning and after pruning.Next,the results are pictured and compared,and the corresponding discussion is given in details.

    Table 1 Parameters of PMSM.

    In Fig.6,the proposed strategy does not use pruning method to check searching tree.The simulation results illustrate that the actual value Iqis close to reference value I?qwith the maximum tracking error less than 0.005A,and Idis changing around?0.03~0.03A.Also,the torque behavior Teis very smooth and tracks the reference torque input well.It shows that the proposed torque control strategy via switching optimization can effectively achieve the motor torque tracking control.But the calculation number of cost function is about 399 when p=3,which is a large computation burden as growth of predictive horizon.In Fig.7,the pruning method is added to the DP solution algorithm.The simulation results illustrate that the actual value Iqcan also track the reference value I?qwell with the maximum tracking error less than 0.007A that is bigger than the case before pruning,and Idis changing around?0.02~?0.02A.Since the proposed method is a model-based optimization algorithm and the PMSM model is a relatively ideal linear model,the model error can cause a small tracking error.The curves of Iqand Idexist slight volatility due to switching optimization.The electromagnetic torque has a similar result as Iqcurve,and the maximum calculation number of cost function is about 22,which is very small.

    In order to analysis the function of pruning method,we compare the results in Table 2.The standard deviation of Iqtracking error is increasing after pruning.Because some optimal control inputs and their branches cannot satisfy the bound function in equation(22),and have possibility to make the closed-loop system unstable,so these control inputs are cut out by pruning method.Meanwhile,the mean calculation amount of cost function is reduced from 399 to about 9.02 when horizon p=3,and from 19607 to about 107.9 when horizon p=5.To verify the computational efficiency of the optimization strategy,we recorded the total computational time for solving the optimization problem.The simulations are running on an eight-core Intel(R)Core(TM)i7 processor with 3.40 GHz processing speed per core and 3.2 GB of RAM available for Win7 32-bit systems.An estimation of CPU time is obtained in MATLAB with ode1(Euler)solver.These results reveal that for simulation time with 3s and sampling time Ts=0.1ms,the CPU time 44.62s of the pruning method is less than time 781.29s before pruning when horizon p=3.It is noted that the computational effort is reduced rapidly in each solution procedure.It shows that the optimization strategy with pruning branch method can still achieve accurate tracking control,and will increase the tracking error,but can greatly reduce the computational burden and decrease the loss of inverter to some extent.This is a trade off between control performance and computational efficiency.

    Fig.6 Simulation results of sinusoidal wave input before pruning.

    Fig.7 Simulation results of sinusoidal wave input after pruning.

    In addition,in the hardware implementation process,to guarantee the switch stability,the switch dead time must be considered.When the single-phase switch signal changes,the switch turn-on command should wait for one dead time after the turn-off command,in order to avoid the single-phase bridge conduction due to shutdown delay,thus ensuring the stability of the inverter.

    Table 2 Constractive analysis of the results.

    5 Conclusions

    In this paper,an MPC-based torque control for hybrid PMSM system is designed to track the torque demand and reduce energy loss.By applying DP algorithm with the pruning branch method,the searching tree for solution is built to look for the optimal control sequence,and the computational burden of input sequence evaluation is reduced in the control scheme.Finally,the simulation evaluates the proposed method can realize the online optimization solution and effectively control the PMSM system to track the torque command and achieve a trade off between satisfying control requirements and reducing computation burden.In addition,there are still some issues waiting to be explored in the future.We need to make an in-depth study on the MPC optimization strategy for hybrid motor system. Aiming at the engineering implementation,the experimental test needs to be developed to verify the real time performance of the proposed controller.

    [1]B.Gasbaoui,A.Nasri.A novel 4WD electric vehicle control strategy based on direct torque control space vector modulation technique.Nonlinear Engineering,2012,3:236–242.

    [2]P.Liu,H.P.Liu.Permanent-magnet synchronous motor drive system for electric vehicles using bidirectional Z-source inverter.IET Electrical Systems in Transportation,2012,4(2):178–185.

    [3]C.H.T.Lee,K.T.Chau,C.H.Liu.Design and analysis of an electronic-geared magnetless machine for electric vehicles.IEEE Transactions on Industrial Electronics,2016,63(11):6705–6714.

    [4]K.C.Kim.A novel calculation method on the current information of vector inverter for interior permanent magnet synchronous motor for electric vehicle.IEEE Transactions on Magnetics,2014,50(2):829–832.

    [5]D.Casadei,F.Profumo,G.Serra.FOC and DTC:two viable schemes for induction motors torque control.IEEE Transactions on Power Electronics,2002,17(5):779–787.

    [6]J.Rodriguez,P.Cortes.Predictive Control of Power Converters and Electrical Drives.Chichester:John Wiley&Sons,2012.

    [7]T.Geyer,G.Papafortiou,M.Morari.Model predictive direct torque contro–Part I:Concept,algorithm,and analysis.IEEE Transactions on Industrial Electronics,2009,56(6):1894–1905.

    [8]C.Bordons,C.Montero.Basic principles of MPC for power converters.IEEE Industrial Electronics Magazine,2015,9(3):31–43.

    [9]P.K.Manakos,T.Geyer,N.Oikonomou,et al.Direct model predictive control:A review of strategies that achieve long prediction intervals for power electronics.IEEE industrial Electronics Magazine,2014,8(1):32–43.

    [10]P.Cortes,M.P.Kazmierkowski,R.M.Kennel,et al.Predictive control in power electronics and drives.IEEE Transactions on Industrial Electronics,2008,55(12):4312–4324.

    [11]F.Morel,X.F.Lin-Shi,J.M.Retif,et al.A comparative study of predictive current control schemes for a permanent-magnet synchronous machine drive.IEEE Transactions on Industrial Electronics,2009,56(7):2715–2728.

    [12]M.Preindl,S.Bolognani.Model predictive direct torque control with finite control set for pmsm drive systems,part 1:Maximum torque per ampere operation.IEEE Transactions on Industrial Informatics,2013,9(4):1912–1921.

    [13]J.J.Hong,D.H.Pan,Z.J.Zong.Comparison of the two current predictive-control methods for a segment-winding permanent-magnetlinear synchronous motor.IEEE Transactions on Plasma Science,2013,41(5):1167–1173.

    [14]F.Barrero,J.Prieto,E.Levi,et al.An enhanced predictive current control method for asymmetrical six-phase motor drives.IEEE Transactions on Industrial Electronics,2011,58(8):3242–3252.

    [15]S.Kouro,P.Corets,R.Vargas,et al.Model predictive controla simple and powerful method to control power converters.IEEE Transactions on Industrial Electronics,2009,56(6):1826–1838.

    [16]G.Prior,M.Krstic.A control Lyapunovapproach to finite control set model predictive control for permanent magnet synchronous motors.ASME Journal of Dynamic Systems,Measurement,and Control,2015,137(1):1–10.

    [17]M.J.Duran,J.Prieto,F.Barrero,et al.Predictive current control of dual three-phase drives using restrained search techques.IEEE Transactions on Industrial Electronics,2011,58(8):3253–3263.

    [18]T.Geyer.Computationally efficient model predictive direct torque control.IEEE Transactions on Power Electronics,2011,26(10):2804–2816.

    [19]D.P.Bertsekas.Dynamic Programming and Optimal Control.3rd ed.Nashua:Athena Scientific,2005.

    [20]D.Graovac,M.P¨urschel.IGBT Power Losses Calculation Using the Data-Sheet Parameters.129th ed.Neubiberg:Infineon Technologies AG,2009.

    [21]H.Chen,F.Allg¨ower.A quasi-infinite horizon nonlinear model predictive control scheme with guaranteed stability.Automatica,1998,34(10):1205–1217.

    [22]H.Chen.Model Predictive Control.1st ed.System and Control Series.Beijing:Science Press,2013.

    [23]J.F.Stumper,A.D¨oltinger,R.Kennel.Classical model predictive control of a permanent magnet synchronous motor.EPE Journal,2012,22(3):24–31.

    [24]E.L.Lawler,D.E.Wood.Branch-and-bound methods:A survey.Operations Research,1996,14(4):699–719.

    26 December 2016;revised 17 February 2017;accepted 17 February 2017

    DOI10.1007/s11768-017-6193-z

    ?Corresponding author.

    E-mail:chenh@jlu.edu.cn.Tel./fax:+86-431-85691900.

    This work was supported by the NSFC Projects of International Cooperation and Exchanges(No.61520106008),the National Natural Science Foundation of China(Nos.61503149,U1564207)and the Graduate Innovation Fund of Jilin University(No.2016093).

    ?2017 South China University of Technology,Academy of Mathematics and Systems Science,CAS,and Springer-Verlag Berlin Heidelberg

    Bingtao RENreceived the B.Sc.degree in Automation from Jilin University,Changchun,China,in 2012,where he is currently working toward the Ph.D.degree in Control Theory and Control Engineering.His current research interests include torque coordination optimization and motor control of electric vehicles.E-mail:tao11235821@163.com.

    Hong CHENreceived the B.Sc.and M.Sc.degrees in Process Control from Zhejiang University,Zhejiang,China,in 1983 and 1986,respectively,and the Ph.D.degree in System Dynamics and Control Engineering from the University of Stuttgart,Stuttgart,Germany,in1997.Since1999,she has been a Professor in Jilin University,Changchun,China,where she currently serves as Tang Aoqing Professor and as the Director of the State Key Laboratory of Automotive Simulation and Control.Her current research interests include model predictive control,optimal and robust control,nonlinear control and applications in mechatronic systems focusing on automotive systems.E-mail:chenh@jlu.edu.cn.

    Haiyan ZHAOreceived the B.Sc.degree in Automation,the M.Sc.and Ph.D.degree in Control Theory and Control Engineering from Jilin University,Changchun,China,in 1998,2004 and 2007,respectively.Since 2007,she has been a lecturer in Jilin University,Changchun,China.Her current research interests include vehicle stability control and state estimation of electric vehicles.E-mail:zhao_hy@jlu.edu.cn.

    Wei XUreceived the B.Sc.degree in Automation from Harbin University of Science and Technology,Harbin,China,in 2010,and received the M.Sc.degree in Pattern Recognition and Intelligent System from Jilin University,Changchun,China,in 2015,where she is currently working toward the Ph.D.degree in Control Theory and Control Engineering.Her current research focuses on regenerative braking system of electric vehicles.E-mail:wei.xu.jlu@hotmail.com.

    国产日韩欧美亚洲二区| 丝瓜视频免费看黄片| 一本久久精品| 亚洲av福利一区| 黑人巨大精品欧美一区二区蜜桃 | 黄色视频在线播放观看不卡| 丝袜脚勾引网站| 午夜福利影视在线免费观看| 高清欧美精品videossex| 高清欧美精品videossex| 97精品久久久久久久久久精品| 秋霞在线观看毛片| 亚洲五月色婷婷综合| 老司机影院成人| 美女国产高潮福利片在线看| 看免费av毛片| 亚洲av.av天堂| 国产女主播在线喷水免费视频网站| 日本黄大片高清| 免费观看在线日韩| 日本黄大片高清| 你懂的网址亚洲精品在线观看| 国产极品天堂在线| √禁漫天堂资源中文www| 毛片一级片免费看久久久久| 夜夜骑夜夜射夜夜干| 亚洲精品456在线播放app| 欧美精品高潮呻吟av久久| 精品酒店卫生间| 日韩av不卡免费在线播放| 日韩精品有码人妻一区| 日日爽夜夜爽网站| 国产精品国产三级国产专区5o| 97在线人人人人妻| 欧美激情极品国产一区二区三区 | 高清欧美精品videossex| 欧美3d第一页| 国产成人午夜福利电影在线观看| 日韩一本色道免费dvd| 国产成人免费观看mmmm| 亚洲 欧美一区二区三区| a级毛片黄视频| 亚洲国产成人一精品久久久| 亚洲欧美成人精品一区二区| 国产成人精品无人区| 色婷婷av一区二区三区视频| 亚洲 欧美一区二区三区| 国产成人欧美| 人人妻人人澡人人爽人人夜夜| 久久久久精品久久久久真实原创| 啦啦啦视频在线资源免费观看| 国产精品久久久久久久久免| 国产白丝娇喘喷水9色精品| 制服诱惑二区| av在线播放精品| 香蕉精品网在线| 深夜精品福利| 久久久久精品久久久久真实原创| 99久久精品国产国产毛片| 国产亚洲av片在线观看秒播厂| 欧美最新免费一区二区三区| 免费不卡的大黄色大毛片视频在线观看| 春色校园在线视频观看| 亚洲国产毛片av蜜桃av| 精品熟女少妇av免费看| 国产男人的电影天堂91| 老司机影院成人| 在线观看免费视频网站a站| 欧美精品一区二区大全| 精品久久国产蜜桃| 少妇人妻精品综合一区二区| 亚洲精品美女久久久久99蜜臀 | av线在线观看网站| 国产有黄有色有爽视频| 少妇的丰满在线观看| 在现免费观看毛片| 国产成人午夜福利电影在线观看| av播播在线观看一区| 精品第一国产精品| 制服丝袜香蕉在线| 午夜激情久久久久久久| 亚洲精品成人av观看孕妇| 久久久欧美国产精品| 1024视频免费在线观看| 欧美日韩视频高清一区二区三区二| 久久99热6这里只有精品| 亚洲国产精品999| 久久99精品国语久久久| 精品少妇内射三级| 日本欧美视频一区| 国产精品 国内视频| www.熟女人妻精品国产 | 欧美人与性动交α欧美精品济南到 | 亚洲色图 男人天堂 中文字幕 | 人妻一区二区av| 亚洲色图 男人天堂 中文字幕 | 尾随美女入室| 观看av在线不卡| 久久97久久精品| 成年av动漫网址| 夫妻性生交免费视频一级片| 国产亚洲欧美精品永久| 午夜激情av网站| 国产成人午夜福利电影在线观看| 亚洲天堂av无毛| 久久97久久精品| 久久精品国产亚洲av天美| 国产高清不卡午夜福利| 亚洲综合精品二区| 极品少妇高潮喷水抽搐| 久久人妻熟女aⅴ| 国产一区亚洲一区在线观看| 午夜激情av网站| 久久亚洲国产成人精品v| 黄片无遮挡物在线观看| 卡戴珊不雅视频在线播放| 国产深夜福利视频在线观看| 亚洲欧美精品自产自拍| 国产高清国产精品国产三级| 高清毛片免费看| 免费观看av网站的网址| 久久ye,这里只有精品| 欧美丝袜亚洲另类| 亚洲中文av在线| 中国三级夫妇交换| 麻豆精品久久久久久蜜桃| 亚洲久久久国产精品| 一区在线观看完整版| 少妇猛男粗大的猛烈进出视频| 老熟女久久久| 日本黄大片高清| 国国产精品蜜臀av免费| 久久久久久久大尺度免费视频| 精品国产一区二区三区久久久樱花| 色婷婷av一区二区三区视频| 亚洲五月色婷婷综合| 不卡视频在线观看欧美| 日韩制服骚丝袜av| 亚洲欧美一区二区三区国产| 精品酒店卫生间| 久久久久视频综合| 亚洲av综合色区一区| 精品酒店卫生间| 国产精品国产三级国产av玫瑰| 亚洲国产精品成人久久小说| 免费看光身美女| 亚洲伊人久久精品综合| 久久 成人 亚洲| 九色亚洲精品在线播放| 国产精品一区二区在线观看99| 成人国语在线视频| 国产69精品久久久久777片| 国产在线免费精品| 黑人猛操日本美女一级片| 久久精品久久久久久久性| 久久久国产精品麻豆| 亚洲国产欧美在线一区| 青春草国产在线视频| 久久鲁丝午夜福利片| 精品一区二区免费观看| 国产日韩欧美视频二区| 亚洲成人av在线免费| 有码 亚洲区| a级毛色黄片| 久久久久久久久久久免费av| 精品久久蜜臀av无| 18禁动态无遮挡网站| 九九在线视频观看精品| 国产视频首页在线观看| 精品亚洲成a人片在线观看| 边亲边吃奶的免费视频| 黄片播放在线免费| 在线观看一区二区三区激情| 欧美日韩视频精品一区| 色婷婷久久久亚洲欧美| 成人亚洲精品一区在线观看| 亚洲欧洲精品一区二区精品久久久 | 亚洲国产av新网站| 黄网站色视频无遮挡免费观看| 大香蕉久久网| 国产亚洲最大av| 三上悠亚av全集在线观看| 亚洲欧美一区二区三区黑人 | 在线观看www视频免费| 久久久久久久久久久免费av| 狂野欧美激情性xxxx在线观看| 狠狠精品人妻久久久久久综合| 欧美日韩一区二区视频在线观看视频在线| 91在线精品国自产拍蜜月| 国产高清三级在线| 成人手机av| 亚洲高清免费不卡视频| 亚洲成av片中文字幕在线观看 | 亚洲欧洲精品一区二区精品久久久 | 少妇的丰满在线观看| a级毛色黄片| 国产av精品麻豆| 精品久久国产蜜桃| 91成人精品电影| 丝袜人妻中文字幕| www日本在线高清视频| 久久亚洲国产成人精品v| 少妇人妻久久综合中文| 成人二区视频| 日韩av在线免费看完整版不卡| 黄色毛片三级朝国网站| 秋霞伦理黄片| 色吧在线观看| 晚上一个人看的免费电影| 欧美日韩一区二区视频在线观看视频在线| 亚洲,一卡二卡三卡| 新久久久久国产一级毛片| 天天躁夜夜躁狠狠久久av| av国产精品久久久久影院| freevideosex欧美| 啦啦啦中文免费视频观看日本| 国产成人精品一,二区| 国产黄频视频在线观看| 九色亚洲精品在线播放| 制服丝袜香蕉在线| 国产免费一区二区三区四区乱码| 国产在线视频一区二区| 九色成人免费人妻av| 亚洲成人一二三区av| 免费女性裸体啪啪无遮挡网站| 成人18禁高潮啪啪吃奶动态图| 久久精品国产鲁丝片午夜精品| 亚洲精品自拍成人| 亚洲婷婷狠狠爱综合网| 国产一区二区三区综合在线观看 | 国产成人一区二区在线| 精品国产一区二区久久| 看免费av毛片| 久久精品夜色国产| 亚洲欧洲精品一区二区精品久久久 | 久久青草综合色| 99热全是精品| 国产伦理片在线播放av一区| 亚洲熟女精品中文字幕| 国产成人免费无遮挡视频| 在线观看人妻少妇| 亚洲欧美一区二区三区黑人 | 亚洲综合色惰| 777米奇影视久久| 高清不卡的av网站| 999精品在线视频| 国产伦理片在线播放av一区| 亚洲精品久久久久久婷婷小说| 亚洲精品国产色婷婷电影| 亚洲人与动物交配视频| 在线观看一区二区三区激情| 伦理电影大哥的女人| 国产一区二区在线观看日韩| 久久综合国产亚洲精品| 最近的中文字幕免费完整| 久久精品aⅴ一区二区三区四区 | av视频免费观看在线观看| 亚洲中文av在线| 中文乱码字字幕精品一区二区三区| 国产爽快片一区二区三区| 国产成人午夜福利电影在线观看| 男女免费视频国产| 我的女老师完整版在线观看| 天美传媒精品一区二区| 韩国av在线不卡| 中国美白少妇内射xxxbb| 香蕉丝袜av| 久久精品国产自在天天线| 免费av中文字幕在线| 中文乱码字字幕精品一区二区三区| 免费黄频网站在线观看国产| 99精国产麻豆久久婷婷| www.色视频.com| 五月天丁香电影| 久久av网站| 日韩中文字幕视频在线看片| 伊人久久国产一区二区| 亚洲欧洲国产日韩| 久热久热在线精品观看| 中文天堂在线官网| 国产精品女同一区二区软件| 2022亚洲国产成人精品| 两性夫妻黄色片 | 丰满乱子伦码专区| 五月天丁香电影| 精品一区二区三卡| 大陆偷拍与自拍| 国产成人免费观看mmmm| 日韩欧美精品免费久久| 久久精品国产鲁丝片午夜精品| 90打野战视频偷拍视频| 黄片无遮挡物在线观看| 中文精品一卡2卡3卡4更新| 男男h啪啪无遮挡| 国产精品一二三区在线看| 亚洲丝袜综合中文字幕| 一级片免费观看大全| 少妇猛男粗大的猛烈进出视频| 熟女av电影| xxx大片免费视频| 亚洲美女搞黄在线观看| 精品国产乱码久久久久久小说| 青春草国产在线视频| 精品福利永久在线观看| 亚洲国产色片| 欧美精品一区二区大全| av女优亚洲男人天堂| 成人亚洲欧美一区二区av| 久久99热这里只频精品6学生| 国产午夜精品一二区理论片| xxx大片免费视频| 97在线人人人人妻| 桃花免费在线播放| 丝瓜视频免费看黄片| 男女边摸边吃奶| 亚洲av福利一区| 人妻系列 视频| 另类亚洲欧美激情| 这个男人来自地球电影免费观看 | 久久久精品免费免费高清| 18在线观看网站| 久久综合国产亚洲精品| 晚上一个人看的免费电影| 亚洲久久久国产精品| 如日韩欧美国产精品一区二区三区| 日本爱情动作片www.在线观看| 超碰97精品在线观看| 亚洲成色77777| 99热网站在线观看| 高清欧美精品videossex| 色94色欧美一区二区| 啦啦啦中文免费视频观看日本| 国产毛片在线视频| 中国国产av一级| 国产又色又爽无遮挡免| 97在线视频观看| tube8黄色片| 久久久久久伊人网av| 欧美成人午夜免费资源| 亚洲av成人精品一二三区| 女人被躁到高潮嗷嗷叫费观| 欧美日韩精品成人综合77777| 人妻系列 视频| 亚洲一区二区三区欧美精品| 欧美成人午夜精品| 亚洲美女黄色视频免费看| 少妇人妻精品综合一区二区| 丝袜人妻中文字幕| 热re99久久精品国产66热6| 日韩一本色道免费dvd| 国产片特级美女逼逼视频| 女性生殖器流出的白浆| 欧美精品国产亚洲| 色婷婷久久久亚洲欧美| 久久毛片免费看一区二区三区| 久久99蜜桃精品久久| 亚洲欧美日韩卡通动漫| av黄色大香蕉| 欧美日韩一区二区视频在线观看视频在线| 国产精品国产三级国产av玫瑰| 97人妻天天添夜夜摸| kizo精华| av免费观看日本| 国产亚洲精品久久久com| 亚洲三级黄色毛片| 国产黄色免费在线视频| 亚洲色图综合在线观看| 在线观看www视频免费| 国产欧美亚洲国产| 99香蕉大伊视频| 一区二区三区乱码不卡18| 中国美白少妇内射xxxbb| 一区二区av电影网| 天堂俺去俺来也www色官网| 久久午夜综合久久蜜桃| 精品一品国产午夜福利视频| 精品一区在线观看国产| 久久亚洲国产成人精品v| 一级a做视频免费观看| 国产色婷婷99| 一边亲一边摸免费视频| 久热这里只有精品99| 欧美国产精品一级二级三级| 亚洲人成网站在线观看播放| 黄片播放在线免费| 大香蕉久久网| 精品国产国语对白av| 免费av不卡在线播放| 好男人视频免费观看在线| 免费av不卡在线播放| videossex国产| 香蕉丝袜av| 久久国内精品自在自线图片| 夫妻性生交免费视频一级片| 国产精品久久久久成人av| 寂寞人妻少妇视频99o| 老司机亚洲免费影院| 日本免费在线观看一区| 亚洲欧美中文字幕日韩二区| 美女大奶头黄色视频| 97人妻天天添夜夜摸| 最新的欧美精品一区二区| 国产精品国产三级国产专区5o| 亚洲精品,欧美精品| 婷婷色av中文字幕| 草草在线视频免费看| 高清在线视频一区二区三区| xxx大片免费视频| 久久婷婷青草| 久久热在线av| 国产伦理片在线播放av一区| 国产免费视频播放在线视频| 你懂的网址亚洲精品在线观看| 亚洲av免费高清在线观看| 国产免费福利视频在线观看| 久久久久国产精品人妻一区二区| 91国产中文字幕| 亚洲精品456在线播放app| 国产一区二区在线观看日韩| av黄色大香蕉| 久久人人爽人人爽人人片va| a级片在线免费高清观看视频| 欧美性感艳星| 女性生殖器流出的白浆| av不卡在线播放| 在线看a的网站| 国产黄色视频一区二区在线观看| 黄片无遮挡物在线观看| av线在线观看网站| 亚洲综合色网址| 久久久欧美国产精品| 青青草视频在线视频观看| 亚洲精品美女久久久久99蜜臀 | 亚洲一码二码三码区别大吗| 国产黄色视频一区二区在线观看| 日本av免费视频播放| 香蕉精品网在线| www.色视频.com| 在线观看一区二区三区激情| 国国产精品蜜臀av免费| 欧美人与善性xxx| 91精品伊人久久大香线蕉| 国产高清不卡午夜福利| a级毛片黄视频| 最新的欧美精品一区二区| 免费看av在线观看网站| 精品久久久久久电影网| 国产色婷婷99| 久久热在线av| 久久女婷五月综合色啪小说| 曰老女人黄片| 制服诱惑二区| 新久久久久国产一级毛片| 九色成人免费人妻av| 国产精品不卡视频一区二区| 久久97久久精品| 色网站视频免费| 99久国产av精品国产电影| 一级,二级,三级黄色视频| 超碰97精品在线观看| 搡女人真爽免费视频火全软件| av在线老鸭窝| 在线观看国产h片| 国产在线免费精品| 国内精品宾馆在线| 一区二区三区精品91| 午夜精品国产一区二区电影| 成人国产av品久久久| 青春草国产在线视频| 国产一区二区在线观看日韩| 亚洲国产成人一精品久久久| 亚洲国产av新网站| 在线观看免费视频网站a站| 黑人欧美特级aaaaaa片| 久久精品人人爽人人爽视色| 亚洲国产精品成人久久小说| 午夜av观看不卡| 十八禁高潮呻吟视频| 一边亲一边摸免费视频| 精品少妇黑人巨大在线播放| 精品国产国语对白av| 永久免费av网站大全| 国产精品一二三区在线看| 又大又黄又爽视频免费| 成人综合一区亚洲| 日韩成人av中文字幕在线观看| 久久久久久久久久人人人人人人| 水蜜桃什么品种好| 国产一区二区在线观看日韩| 夜夜爽夜夜爽视频| 两个人看的免费小视频| 欧美日韩综合久久久久久| 下体分泌物呈黄色| 欧美人与性动交α欧美精品济南到 | 欧美3d第一页| 一级毛片 在线播放| 午夜福利乱码中文字幕| 高清不卡的av网站| 男女免费视频国产| 国产精品久久久久久精品古装| 亚洲伊人久久精品综合| av视频免费观看在线观看| 久久久久久久精品精品| 一级毛片黄色毛片免费观看视频| 亚洲五月色婷婷综合| 日韩精品有码人妻一区| 日韩一区二区视频免费看| 日韩成人av中文字幕在线观看| 精品国产一区二区三区久久久樱花| 欧美xxxx性猛交bbbb| 中文字幕人妻熟女乱码| 精品国产一区二区三区四区第35| 黄片播放在线免费| 香蕉丝袜av| 日韩一区二区三区影片| 桃花免费在线播放| 97超碰精品成人国产| 99久久综合免费| 日韩一本色道免费dvd| 午夜精品国产一区二区电影| 少妇的逼好多水| 国产欧美亚洲国产| 1024视频免费在线观看| 久久久久久久大尺度免费视频| 日韩三级伦理在线观看| 亚洲在久久综合| 国产精品久久久久成人av| 只有这里有精品99| 80岁老熟妇乱子伦牲交| 精品少妇内射三级| 亚洲av国产av综合av卡| 国产爽快片一区二区三区| 国产激情久久老熟女| 十八禁网站网址无遮挡| 精品一区二区三区视频在线| 99香蕉大伊视频| 免费人成在线观看视频色| 亚洲精品久久午夜乱码| 男女免费视频国产| 中文字幕制服av| 日韩av免费高清视频| 十八禁网站网址无遮挡| 美女国产高潮福利片在线看| 欧美bdsm另类| 成年动漫av网址| 午夜福利视频精品| 天天躁夜夜躁狠狠久久av| 自线自在国产av| 欧美少妇被猛烈插入视频| 黄色配什么色好看| 国产 精品1| 免费播放大片免费观看视频在线观看| 日韩欧美一区视频在线观看| 精品少妇内射三级| 婷婷色麻豆天堂久久| 国产深夜福利视频在线观看| 国产精品一区二区在线不卡| 最近中文字幕2019免费版| 免费黄色在线免费观看| 一级毛片 在线播放| 国产成人一区二区在线| 亚洲av免费高清在线观看| 欧美日韩精品成人综合77777| 男女无遮挡免费网站观看| 一区二区三区四区激情视频| 亚洲精品aⅴ在线观看| 18禁观看日本| 91久久精品国产一区二区三区| 精品国产一区二区三区久久久樱花| a级毛色黄片| 少妇的丰满在线观看| 在线天堂中文资源库| 国产 精品1| 国产成人91sexporn| kizo精华| 国产精品久久久久成人av| 69精品国产乱码久久久| 美女中出高潮动态图| 久久韩国三级中文字幕| 男女边摸边吃奶| 色5月婷婷丁香| 高清不卡的av网站| 中文字幕另类日韩欧美亚洲嫩草| www日本在线高清视频| 18在线观看网站| 国产色爽女视频免费观看| xxxhd国产人妻xxx| 日韩av不卡免费在线播放| 极品少妇高潮喷水抽搐| 青青草视频在线视频观看| 边亲边吃奶的免费视频| 国产深夜福利视频在线观看| 亚洲国产精品一区三区| 成人毛片60女人毛片免费| 韩国精品一区二区三区 | 女性生殖器流出的白浆| 国产乱来视频区| 一本—道久久a久久精品蜜桃钙片| 国产在线视频一区二区| 欧美精品亚洲一区二区| www.色视频.com| 91精品三级在线观看| 日韩 亚洲 欧美在线| 亚洲av.av天堂| 欧美激情国产日韩精品一区| 99久国产av精品国产电影| 制服诱惑二区| 两个人免费观看高清视频| 亚洲欧美清纯卡通| 亚洲熟女精品中文字幕| 人妻 亚洲 视频| 丝袜美足系列| 亚洲av国产av综合av卡| 老女人水多毛片| 在线观看美女被高潮喷水网站|