• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    For an easily implementable method for model predictive control of a DC / DC converter, and a corresponding controller with which the optimization problem of the model predictive control can be solved sufficiently fast even with large prediction horizons, it is provided that the optimization problem is divided into two optimization problems by using in the control unit (10 a model predictive output control and a model predictive inductor current control are implemented, wherein for output control the strings of the multi-phase DC / DC converter (12) are combined into a single string and a time discrete state space model is computed therefrom, and output control is based on a first cost function (Jv) of the output control optimization problem predicts the input voltage (uv, k + 1) of the next sampling step (k + 1) for this single skew, which is set as default to the inductor current control and the D Rosselstromregelung therefrom on the basis of a second cost function (Ji) of the optimization problem of the inductor current control for the next scanning step (k + 1) the necessary switch positions of the switches (S1, S2, S3, S4, S5, S6) of the strands of the multi-phase DC / DC converter (12) determined.
    公开号:AT513776A2
    申请号:T50265/2014
    申请日:2014-04-08
    公开日:2014-07-15
    发明作者:
    申请人:Avl List Gmbh；
    IPC主号:
    专利说明:

    AV-3598 AT
    Method and controller for model-predictive control of a multiphase DC / DC
    converter
    The present invention relates to a method for model predictive control and a model predictive controller of a multiphase DC / DC converter with a half-bridge with two switches for each strand of the polyphase DC / DC converter, wherein the switches are driven to generate a desired output from a control unit ,
    A battery emulator is known to mimic the behavior of an electrical battery. Such battery emulators 1 are e.g. in the development or testing of powertrains of electric vehicles or hybrid vehicles or for the development of electrical energy storage required such vehicles, as indicated in Figure 1. A battery emulator 1 usually generates a DC output voltage v2 as a function of a load current i2. For this purpose, the load current i2 is measured and supplied to a battery model 4, which calculates a reference output voltage v2R from the load current i2, which is then generated by the battery emulator 1 at its DC voltage output. At the battery emulator 1 any, real electrical load 5 is connected. For example, in Fig. 1 is formed of an inverter 2 which drives an electric motor M, which in turn drives a mechanical load ML (e.g., a vehicle). The inverter 2 may also be connected to electrical loads EL, e.g. electrical components of a vehicle (for example, an air conditioner, sound system, lighting system, etc.). The battery emulator 1 and the load 5 or an inverter 5 of the load 5 can be controlled by a control unit 3, e.g. an ECU of the vehicle to be controlled.
    A battery tester 7 is known to be used to load a real electric battery 6 by a certain load, in the form of a DC load current i2, as shown in Fig.1a. Depending on the state of the battery 6 (SoC, SoH), a specific DC output voltage v2 sets in which can be measured. For testing a battery 6, e.g. As part of the battery development, 6 predetermined test runs, in the form of a predetermined time course of the DC load current i2, performed with the battery. The battery tester 7 can be controlled by a control unit 3 for this purpose.
    Power electronics in the form of a DC / DC converter, which generates and provides the required output voltage v2 (battery emulator 1) or the required output current i2 (battery tester 7), is typically implemented in the battery emulator 1 or battery tester 7 for this purpose. The battery emulator 1 or the battery tester 7 can be powered by one, usually a 3-phase, AC voltage source AC, which is rectified internally, or by a DC voltage source. Such a battery emulator 1 is exemplified 2/261
    AV-3598 AT shown in Fig.2. In the case of a battery tester 7, an additional inductor L2 can also be arranged on the output side, otherwise the circuits match, as shown in FIG. 2a. On the input side, the three-phase AC voltage AC is rectified in a rectifier 11 and a smoothing capacitor C0 to a DC voltage V0. Connected thereto are a three-phase DC / DC converter 12 with parallel half-bridges and inductors La, Lb, Lc, which are each driven by a half-bridge, and an output-side smoothing capacitor C or a further inductor L2 (in the case of a battery tester 7). Such DC / DC converters 12 are well known, which is why will not be discussed in detail here. The switches S1... S6 of the half bridges are driven by a control unit 10 in order to set the desired output voltage v2 or the desired output current i2. In the control unit 10, a well-known pulse width modulation (PWM) is usually provided in order to set the output voltage v2 or the desired output current i2 via the duty cycle of the switches. In a PWM, the switches are switched once at each sampling instant, given a certain sampling rate. The sampling rate is thus dependent on the permissible frequency with which the switches S1... S6, as a rule, insulated gate bipolar transistors (IGBTs) or metal oxide semiconductor field effect transistors (MOSFETs) can be switched. However, the frequency with which the switches can be switched is limited by the switching losses arising during switching. Since the PWM switches at each sampling step, this limitation also limits the sampling rate and thus also the controller bandwidth. This limitation leads to poor dynamics of the regulation of such transducers 12, in that it is only possible to respond slowly to disturbances or transient switching processes of the load 5. Although an increase of the sampling rate in the form of oversampling is possible, but only under severe limitations, oversampling for the regulation of the DC / DC converter 12 has no practical relevance.
    In order to avoid this disadvantage of a PWM, a new control strategy, the so-called Finite Control Set Model Predictive Control (FCS-MPC) has already been introduced. In this control strategy, the switches S1 ... S6 are directly controlled, which is why a PWM can be dispensed with. Thus, the sampling rate can be increased and the dynamics of the control can be improved. Such methods for direct drive of the switches in power electronic systems are not new. An overview of this can e.g. in J. Rodriguez, et al., "State of the art of finite control set model predictive control in power electronics", Industrial Informatics, IEEE Transactions, 9 (2): 1003-1016, May 2013. In EP 2 528 225 B1 this control strategy finds e.g. Application for the regulation of an electrical machine. 3/262,
    AV-3598 AT FCS-MPC is characterized by the limited number of possibilities for the manipulated variable, the so-called finite control set. For the switches S1... S6 of the half bridges of the DC / DC converter 12 in FIG. 2 or FIG. 2a, the necessary condition is that in each half bridge both switches are never open or closed at the same time 8 (23) possible switch positions that make up the finite control set. Methods of model predictive control are known to be based on an optimization problem in the form of minimizing a quality function, also called a cost function. The problem here is that a switching sequence according to the selected prediction horizon (ie a forecast of the future switch positions) is included in the quality function. Thus, the optimization problem grows exponentially with the prediction horizon. Prediction Horizon is the number of sampling steps that will be considered in the future. With 8 possible switch settings, with a prediction horizon of 1 81 = 8, there are 8 possibilities for which the cost function must be solved in order to find the optimum of these possibilities. With a prediction horizon of 5, however, 85 = 32,000 solutions would already result, and with a prediction horizon of 10 already over one billion solutions would result. However, the solution of the optimization problem must be found in a very short time for a desired real-time control. If at a sampling rate of e.g. 20kHz is sampled, then the solution must be within one sampling step, ie within 50 ps. Even with the currently available very fast processors, this can not be done beyond a certain prediction horizon. For the control of DC / DC converters with FCS-MPC, however, large prediction horizons (> 10) are sought in order to avoid undesired overshoot in transient control processes, e.g. at load jumps, reduce.
    WO 2013/174967 A1 describes a model-predictive control method for a battery emulator and in WO 2013/174972 A1 a model-predictive control method for a battery tester. This generally explains the method of model predictive control and specifies a method with which the optimization problem can be solved sufficiently quickly to enable sampling rates in the kHz range. The control of the DC / DC converter takes place here but again by means of a PWM with all the disadvantages explained above, in particular the limitation of the sampling rate.
    The subject invention has now set itself the goal of providing an easily implementable method for model-predictive control of a DC / DC converter, and a corresponding controller, with which the optimization problem can be solved sufficiently quickly, even with large prediction horizons.
    This object is achieved according to the invention by dividing the optimization problem of the model-predictive control into two optimization problems by specifying in FIG. 4/263 '.
    AV-3598 AT
    A model predictive output regulator and a model predictive inductor current regulator are implemented, wherein the outputs of the multistage DC / DC converter are combined into a single strand and a time discrete state space model is created and the output variable controller based on a first cost function of the optimization of the output variable controller, the input voltage of next sampling step predicts for this single strand, which is the throttle current controller as a default and the throttle current controller determines the necessary switch positions of the switches of the strands of the polyphase DC / DC converter based on a second cost function of the optimization problem of the inductor current regulator for the next sampling. By dividing the controller according to the invention into two model-predictive, cascaded sub-controllers, the output variable controller and the inductor current controller, it is possible to reduce the state space model from a fourth-order model to a second-order model. This also reduces the finite control set of the model predictive control, which significantly reduces the solution space for the optimization problem. In particular, because the computational effort for the inductor current control compared to the computational effort of the output variable control can be neglected, since the degree of freedom of the inductor current control is very much reduced. Solutions of the optimization problem can therefore be found more quickly with the inventive approach, which makes it possible to use even larger prediction horizons with high sampling rates.
    The potential solution space of the optimization problem can still be significantly reduced if the solution space is investigated in advance in the form of a set of possible solutions to the optimization problem of the output variable controller and solutions that can not occur are eliminated from the solution space. This can advantageously take place in such a way that a simulation run is carried out with a predetermined control sequence of the output variable and the resulting input vectors of the DC / DC converter are recorded and the input voltage of the output variable control is reconstructed from the input vectors and the occurring sequences of the input voltage are reduced Solution space are stored.
    Likewise, if solutions to the output controller's optimization problem are represented as an integer linear combination of basic functions, the base functions for the first sampling steps have a width of one sampling step and the base functions for the subsequent sampling steps a width of an integer multiple of one Have scanning step. -4- 5/26
    AV-3598 AT
    A further limitation of the solution space can be achieved if the integer linear combinations are subject to the restriction that solutions from one sample step to the next sample step must not switch more than one.
    The subject invention will be explained in more detail below with reference to Figures 1 to 13, which show by way of example, schematically and not by way of limitation advantageous embodiments of the invention. It shows
    1 shows a known test arrangement for testing a load with a battery emulator, FIG. 1 a shows a known test arrangement for testing a battery means of a battery tester,
    2 shows the known circuit of a power electronics of a battery emulator,
    2a shows the known circuit of a power electronics of a battery tester,
    3 shows a simplified electrical circuit diagram of a DC / DC converter,
    4 shows a model predictive control according to the invention of a DC / DC converter,
    5 is an electrical model equivalent circuit diagram for the output control of the DC / DC converter,
    6 is an electrical equivalent circuit diagram for the inductor current control of the DC / DC converter,
    7 shows a control scheme according to the invention of a battery emulator,
    8 shows an inventive control scheme of a battery emulator with an observer,
    9 shows the representation of the solutions of the optimization problem as a search tree,
    10b, the activation of the switches of the half-bridge of the DC / DC converter by the inductor current control at a predetermined input (Fig.10a),
    11 shows an exemplary control sequence for the DC / DC converter, and FIGS. 12 and 13 show a method according to the invention for reducing the solution space of the optimization problem.
    Starting point of the method according to the invention is the known model of the battery emulator 1 as shown in Figure 2, or the known model of the battery tester 7 as shown in Figure 2a. The smoothing capacitor C0 is assumed to be sufficiently large, whereby the dynamics of the rectifier 11 is neglected and the DC voltage V0 can be assumed constant. Stray inductances of the cables and coils, non-linearities and parasitic capacitances of the semiconductor switches of the half-bridges are also neglected, which is permissible for normal operating conditions. The switches S1 and S2, S3 and S4, S5 and S6 of the individual bridge arms are always connected in opposition, ie S1 closed and S2 open or vice versa, etc., in order to avoid a short circuit across the smoothing capacitor Co. By appropriate circuit of the switches S1 to S6 can be 6/26
    AV-3598 AT positive and negative currents, ie a bidirectional DC / DC converter 12 realize. Of course, the following also applies to a unidirectional DC / DC converter in which a switch can be replaced by a diode in each half-bridge.
    Although the invention is described below using the example of a battery emulator 1, it should be noted here that the invention is applied in the same way to a battery tester 7 with essentially the same circuit (see FIG. 2a). Any differences will be explained at the appropriate place in the description below.
    These assumptions lead to the simplified model of the battery emulator 1 or in general a multiphase DC / DC converter 12, as shown in FIG. 3, the ohm-10 see resistances Ra, Rb, Rc of the chokes La, Lb, in FIG. Lc are not shown for the sake of simplicity. Also indicated in FIG. 3 is the frequently encountered additional output-side throttle L2 in the case of a battery tester 7.
    The input voltages ua, ub, uc of the multi-phase DC / DC converter 12 are defined by ua = SaV0, ub = SbV0 and uc = ScVo, where the signals Sa, Sb, Sc can assume eight different states, according to the table 1. This results in eight possible states for the system input, which form the finite control set of the system.
    Sa Sb Sc 1 1 1 1 1 0 1 0 1 1 0 0 0 1 1 0 1 0 0 0 1 0 0 0
    Table 1
    The behavior of the load 5 of the DC / DC converter 12 is assumed to be unknown, since the control method should work for a wide range of possible applications of the DC / DC converter 12 20. As load 5, therefore, a constant power load (CPL) is assumed. The power P as product of load current iP and voltage v2 at the input of the load is constant, P = iP-v2 = const. The power requirement in the form of the power P is specified by the load 5. The dependence of the load current iP (= i2) on the output voltage v2 thus represents a non-linearity for the system. An ideal voltage source can also be used as the load model, in particular in the case of a battery tester7, including an impedance model, or another model of the battery could be used. -6- 7/26
    AV-3598 AT
    From Figure 3, the state space model of the DC / DC converter 12 with load 5 can then be derived. The system equations result from the application of Kirchhoff's laws, Ohm's law, as well as the differential equations for an ideal capacitor and an ideal inductance to the circuit of Fig. 3 to f4 dt dih - = ~ -ib - v2 + - ub dt Lbb Lb Lbb di
    R 1 = --- i "- + - 1 dt Lc L Le 4 Le dur 1 /. , , - = - Vn + h + h ~ lO) dt CVfl 6 c 2) _d_ dv. iP ~ iP η +
    The non-linear load current iP is linearized by a current operating point Po = ip, o'V2, o in the form (v2-v20) = 2ip 0-^ -v2. This leads to the state space vector 2.0 gp xm = [ia ib ic V2] T, the input vector um = [ua ub uc] T and the output vector ym with zm = ip, 0 to the time-continuous state space model K 0 0 1 La 4 0 Rb 0 1 Lb 4 0 0 Rc 1 Lc 4 1 1 1 gp cccc An [o 0 0 Ψ »v cm 1 - 00 4 0 0 - 0 0 4 H ~ l · m 0 Λ 1 2 0 0 - 4" c , 0 0 Em V- Bm
    The time-continuous state space model is converted by discretization by means of zero-order hold (ZOH) with the sampling time Ts into a time-discrete state space model (denoted by the subscript d) (which is well known) and at the same time the state vector is supplemented by zm = ip, 0. The state vector x0 then becomes x0 = [ia ib ic v2 ip, o] T and 15 the time-discrete state space model with the input vector u0 = [ua ub uc] T to 8/267 '
    AV-3598 AT
    Am, d Em, d *., * + Bm, d 0 1 0 [o 0 0 1 ° ka k denotes the respective sampling step.
    When using other load models, the time-discrete state space model can also deviate from the one described above without changing the basic principle of the invention. In particular, a time-discrete state space model can also be created analogously for the battery tester 7 according to FIG. For model-predictive control, a cost function J to be optimized is required. For the regulation of a DC / DC converter 12, a cost function is set up for this, which evaluates how well the output voltage v2, or the output current i2, of the DC / DC converter 12 corresponds to a predetermined reference signal Rs, e.g. in the form of a predetermined voltage vPR or a predetermined current iPR, can follow. The cost function should primarily evaluate the deviation of the system output Y0 = [y0, k + iy0, k + 2y0, k + NP] T from the reference signal Rs. In the battery emulator 1, the system output yQ is the output voltage v2 and the battery tester 7 is the output current i2. NP denotes the prediction horizon, which indicates how many time steps k are calculated in the future. Furthermore, the switching losses when switching the switches S1 ... S6 are to be evaluated. It is also advantageous if it is ensured that the phase currents ia, ib, ic of the multiphase DC / DC converter 12 are as equal as possible in order to avoid excessive strand currents flowing through individual strings, which could damage the hardware. It will therefore be a cost function J in the
    Form J = (RS-Y0 J (Rs-Υ0) + λ5 ΤτΤ + λΒ · AC2 formulated.
    Therein, the term (Rs-Y0) r (Rs-Y0) evaluates the deviation of the system output Y0 from the reference signal. With the term λ3 · ΤτΤ the switching losses are evaluated, where T = [tk tk + i tk + Np-i] T with tk = | ua, k-ua, k-i | + | ub, k-Ub, k-i | + | uc, k-uc, k-i | evaluates the frequency of the switching operations of the switches S1... S6, in which the difference between an input quantity and the input quantity from the preceding time step is punished. With the weighting factor As, with which this term is weighted in the cost function J, one obtains an additional tuning parameter for the control. With the term λΒ AC2 the deviation of the
    np
    String currents are evaluated to each other, with AC = ^ max (iakJbkJck) -rmn (iak, ibk, ick). Thus k = 1, the deviation between the maximum phase current and the minimum phase current is punished. λΒ is again a weighting factor which gives the control an additional 9/268 '
    AV-3598 AT chen tuning parameters there. Alternatively, too
    Np AC = Σ fc, * "h, k f + {h, k - ic, k f + L · - iajt f are used. k =
    With the weighting factor As, the ripple of the output voltage v2 or of the output current i2 and the switching frequency of the converter 12 is influenced. The higher the weighting factor As, the greater the ripple and the lower the switching frequency. The weighting factor AB influences the width of the band within which the phase currents fluctuate.
    Of course, a weighting factor for the first term in the above cost function J could also be set. Likewise, the cost function could include other or additional terms to evaluate certain other or further aspects.
    In model-predictive control, boundary conditions can also be taken into account, which represents a particular strength of the model-predictive control. An important constraint for protecting the transducer 12 and preventing the saturation of the chokes La, Lb, Lc is a limitation of the strand currents ia, ib, ic, e.g. in the form imin < ia + ib + ic < imax. The consideration of the constraint may e.g. such that a solution of the cost function J is set to the value infinite when a constraint is violated.
    In a conventional model predictive control, this cost function J would now be minimized for all possible combinations of the input vector u0 = [ua ub uc] T for the prediction horizon NP taking into account the boundary conditions. This yields the input quantities u0, k + i, u0, k + 2 > u0, k + Nc for the next time steps, where Nc denotes the control horizon, which is often equal to the prediction horizon NP. According to the receding horizon principle, only the first control process u0, k + i is applied at any one time, and the remainder is discarded. This is repeated for each time step. This basic principle of a model-predictive control is well known, which is why it will not be discussed in more detail here.
    The model-predictive control can be represented as shown in FIG. The current phase currents ia, k, ib, k > ic, k > the current output voltage v2, k, or the output current i2, k at a battery tester 7, and the current load current iP, k are measured and the control unit 10 is supplied. Therein, the cost function J is minimized as explained above, and the required switch positions of the switches S1, ..., are obtained directly from the determined input vector u0, k + i = [ua, k + iub, k + iuc, k + i] T. S6, which will then be switched in the next time step. 10/2
    AV-3598 AT
    As already stated, however, a very large number of possible combinations of successive input vectors u0, k must be calculated for the optimization of the cost function J, which requires a great deal of computing time. In the present example, for a prediction horizon, we would get NP = 10 at 810 = 1,073,741,824 possible combinations.
    By the method according to the invention described below, the number of possible combinations should be significantly reduced.
    For this purpose, the regulation of the DC / DC converter 12 is divided into an output variable regulation, that is, a voltage regulation of the output voltage u2 in a battery emulator 1 (FIG. 5) or a current regulation of the output current i2 in a battery tester 7, and a reactor current control (FIG , For the output variable regulation, the inductances of the chokes La, Lb, Lc of the individual strings of the DC / DC converter 12 are combined to form an inductance L = LJ3 = Lb / 3 = LJ3 (under the permissible assumption of identical chokes) (FIG. 5). The current h is the sum of the individual inductor currents, ii = ia + ib + ic- Likewise, the individual resistors Ra, Rb, Rc (not shown) become a resistor R = Ra / 3 = Rb / 3 = Rc / 3 ( under the permissible assumption of equal resistance). This reduces the fourth-order model of FIG. 3 to a second-order model according to FIG. 5, and at the same time the possible inputs according to the possible switch positions are also applied to the input voltage of the output variable control uv = {0, Vo / 3, 2Vo / 3, V0 } limited. Thus, the finite control set of the output control consists of only four elements. Analogously to that described with reference to FIG. 3, the time-discrete state space model in the form is again obtained
    Au, d EU, d Xv, k + BtJ, d XvMl = 0 1 0 Av Bv Xi II ΤΞΓ > X o 'H with the state vector xv = [H v2 ip.o] T- For this purpose, a cost function Jv is set up, in which again the deviation of the system output from a reference signal Rs and the
    Switching frequency is evaluated. A term (Rs-YVJ (RS-7v), with Yv = [yv, k + iyv, k + 2yv, k + NP] T, evaluates the deviation and a term λν - (Tvk - TvkAJ (Tvk -TvkA ), with Tv, k = [uv, k uv, k + i ... uv, k + Np-i] T, punctures the change of two temporally successive input variables (input voltage uv of the output variable control), where the weighting factor λν again a tuning parameter the cost function is then -10- 11/26
    AV-3598 AT
    Jv = {RS-YV) T {RS -Yv) + K- & k-Tv ^ J (Tvk-TvJl_,). The boundary condition is analogous to the above to imin ^ ii ^ ima *.
    In the case of a battery tester 7 with current regulation, the state vector would be xv = [ii.sub.Vi.sub.i2.sub.v2] T and analogously also a time-discrete state space model and a cost function Jv are obtained.
    Since the finite control set for the output variable control only consists of four elements, only 4Np combination options result for the solution of the optimization problem. With a prediction horizon NP = 10, this results in only 410 = 1,048,576, which is a factor of 1,000 less than in FIG. 3, combination possibilities.
    The inductor current control according to FIG. 6 provides, inter alia, the state variable ii = ia + ib + ic, which requires the output variable regulation. As described above with reference to FIG. 3, the individual phase currents ia, ib, ic should again be kept within a narrow band, and excessive deviations between the individual phase currents should be avoided. The system model of the inductor current control can therefore, as shown in FIG. 6, be formed by the three strings of the DC / DC converter 12 and an ideal voltage source Uz. The course of the remaining state variables is already predicted by the output variable control. Therefore, in each sampling step k, the predicted voltage at the capacitor C is assigned to the ideal voltage source Uz, ie UZ = V2.
    With the state vector Xj = [ia ib ic uz] T and the input vector u, = [ua Ub uc] T, the time-continuous state space model of the inductor current control according to FIG. 6 results in Γ R 1 " '1 a 0 0 0 0 La ~ X X Rh 1 1 0 0 0 0 0 h ~ x * / + X R 1 1 0 0 c 0 0 Lc ~ x X 0 0 0 0 0 0 0 "V" -V " A {B {y, = [, 1 1 θ] Λ-.
    Cf
    The time-continuous state space model is again discretized. So that the specification of the output variable regulation ii = ia + ib + ic can be met, the condition uv = 1/3 (ua + Ub + uc) must be satisfied, as can be shown after simple derivation. This gives -11- 12/26
    AV-3598 AT from the assumption L = La / 3 = LJ3 = LJ3. A cost function J, which evaluates the switching frequency and the equality of the phase currents, is set up analogously to FIG.
    Jt = XrTTT + Xb-AC2 5
    A secondary condition is - (uak + ubk + uck) = uvk, where uv, k is specified by the output variable control as the predicted voltage of the next sampling step k + 1. Due to this constraint, the inductor current controller always supplies the requested current.
    The inductor current control receives the predicted next input voltage uVik + 1 of the output variable regulation as the default and from this specification the actual input to the system is calculated in the form of the input vector Ui = [ua ub uc] T. The choke-10 current control thus determines the switch positions of the switches S1 ... S6 for the next sampling step k + 1. For the subordinate inductor current control, a prediction horizon NP of 1 suffices. However, the inductor current control must follow the output control, i. if the latter outputs 0 or 3 Vo / 3, the choke current control must output [0 0 0] or [111]. This results in a maximum of three remaining possibilities for the case 1-Vo / 3 or 2-Vo / 3 (e.g., for 2-Vo / 3 [1 1 0], [0 1 1] and [1 0 1]). Thus, the degree of freedom of the inductor current control is fixed. The inductor current controller divides the specification of the output variable regulation onto the strings of the multiphase DC / DC converter 12.
    By the solution of the optimization problem with the cost function J, within the degree of freedom of the inductor current regulation, the optimum voltage vector u.sub.i k is determined, which directly determines the switch positions of the switches S1... S6 of the multiphase DC / DC converter.
    Converter 12 delivers. If all or no string of the multi-phase DC / DC converter 12 is switched, no optimization is necessary and u, = [Vo V0 V0] or Ui = [0 0 0] is applied. If only one or two of the strands are connected, then each result in three different possible switching combinations. Thus, for example, in the case of a switched 25 string, Ui = [V0 0 0], Ui = [0 V0 0] or Ui = [0 0 V0].
    The advantage arises from the fact that the optimization problem is divided into two separate optimization problems, whereby the optimization problem of the inductor current control is negligible for the performance. In simulations it was found that for any prediction horizon NP of the output variable control, a prediction horizon of NP = 1 of the inductor current control is sufficient. This means that for the sought solution of the optimization problem of the inductor current control a maximum of three possible combinations have to be investigated. Thus, the computational effort for the throttle current control over the computational effort of the output control can be neglected. -12- 13/26
    AV-3598 AT For the solution of the optimization problem for the output variable regulation and the choke current control only 4Np possible combinations result.
    FIG. 7 shows the resulting control scheme on the basis of the example of a battery emulator 1 as a block diagram. In the control unit 10, the controller 18 is implemented with the output control 15 for the output control 15 and the throttle current regulator 16 for the throttle current control. The output variable controller 15 receives as input the DC voltage V0 and the reference signal Rs, which is to follow the DC / DC converter 12 at the output. The current state variables x0, k, ie phase currents ia, k, ib.k, ic.k, output voltage v2, k and load current iP, k, are measured and supplied to the control unit 10. From this, the input voltages to be applied for the next sampling step k + 1, among others, k + ii Ub, k + i, uc, k + ii or the switch positions of the switches S1... S6, are determined and sent to the DC / DC converter 12 created.
    If the state vector x can not be measured at all or only partially, then a control-technical observer 17, e.g. in the form of a Kalman filter, to estimate the required state vector xo k from measured quantities z0, k, as shown in FIG. After such observers and their interpretation are well known, this will not be explained here.
    The solutions of the optimization problem can also be represented as a search tree 20 with nodes 21 and leaves 22 and with a depth corresponding to the prediction horizon Np, as shown in FIG. Each leaf 22 of the search tree 20 represents a solution to the optimization problem. The search tree 20 can now be searched completely to solve the optimization problem. However, tree search algorithms can also be used in order to obtain the solution of the optimization problem more quickly. One possible algorithm would be e.g. would be e.g. a well-known Branch & Bound algorithm. Thus, the set of possible solutions is subdivided into sub-sets and suboptimal solutions are recognized by means of barriers and rejected. In the worst case you end up here with a complete search of the search tree 20. After the algorithm is well known, will be omitted here to a more detailed description.
    In order to further reduce the number of possible solutions of 4Np, one can also try to reduce the possible solution space (ie the leaves 22 of the search tree 20) before searching through the search tree 20. Here one can take advantage of the knowledge of the underlying optimization problem.
    The vast majority of possible solutions to the optimization problem are never used. This is because the controller in the control unit 10 in the cost functions J, -13- 14/26
    AV-3598 AT and Jv penalize the switching between states to keep the switching frequency low. As a result, the controller in the control unit 10 does not effect a switching operation of the switches S1 to S6 at any time. However, this gives rise to possibilities of limiting the number of possible solutions, the solution space. 5 By penalizing switching operations in the cost functions, the resulting input signal may look like a steady state operation, such as in shown in Fig.10. Fig. 10 a) shows what the output control specifies and Fig.10 b), as the throttle current control to match the switches S1 ... S6 activated or deactivated. If, in a first method for solution space reduction (hereinafter called PP (principle pattern)), a snapshot of the subsequent NP (prediction horizon) sampling steps k + Np is made for any given time k, and the same is done for all other times, it can be seen that the number of different sequences of the input voltage and the output variable regulation occurring is much lower than the 4NP possibilities. These sequences can be stored and summarized and the control unit 10 as a reduced ιοί 5 sungsraum (search tree 20 with fewer leaves 22) are given. Thus, the number of possible solutions is severely limited. As a result, the above-mentioned tree search algorithms only have to search through a much smaller, rather shattered search tree 20. Here, all remaining leaves 22 of the search tree 20 are searched best because the Branch & Bound algorithm would take too long to prepare (Over-20 head).
    The same applies to a battery tester 7 with the output current i2 as the output analog.
    In order to arrive at these sequences, a simulation run can be performed with a control sequence of the output variable (output voltage v2 at the battery emulator 1, as in FIG. 11, 25 or output current i2 at the battery tester 7), which essentially covers all relevant operating points of the DC / DC converter 12 images. Such a control sequence is shown by way of example in FIG. 11. The occurring input vectors Ui = [ua ub uc] T of the DC / DC converter 12 are recorded and the input uv of the output variable control reconstructed therefrom. Thus, these sequences can be extracted.
    A second method for solution space reduction (called CBF (constraint basis functions hereinafter)) is based on a decomposition of the solution space. The entire solution space considered so far has a base as shown in FIG. 12 or, in other words, u = -U- with uint = {0,1,2,3}. Each solution represents an integer linear combination of these basic functions. Since the controller in the control unit 10 is typically not -14- 15/26 30
    AV-3598 AT switches every sampling step k, see e.g. Fig. 10, one may also choose another base, e.g. shown in Fig13. The controller in the control unit 10 has full freedom for the first two scanning steps and is subsequently resolved to a coarser resolution, here in the form of basic functions with a width of three scanning steps. The solution space is now restricted to the integer linear combination of these basis functions.
    Of course, other basic functions than those shown can be used, as long as the solution space is reduced. For example, it would be conceivable to make the temporal resolution even coarser for even larger prediction horizons
    Furthermore, the restriction may be introduced that the solutions may not switch more than one from one scanning step to the next, e.g. 1 Vo / 3 to 2 Vo / 3, but not to 3-Vo / 3. This leads to a further reduction of the solution space by several orders of magnitude. For the implementation, all of these solutions can be precalculated and made available to the control unit 10 as a possible solution space. For this purpose, all possible integer linear combinations are evaluated offline and those are excluded that violate the above restriction. The remaining linear combinations are then available to the online algorithm as a reduced search space. This reduced solution space can now be passed to a tree search algorithm in order to search the reduced search tree 20 as effectively as possible. Due to the structure of the resulting search tree 20, Branch & BoundAlgorithm very good. The same applies analogously to a battery tester 7 with the output current i2 as the output variable.
    Table 2 shows the effect of the methods of reducing the solution space of the optimization problem described above. From this it can be seen that the number of possible solutions to the optimization problem can be massively reduced with the above methods, which also makes it possible to calculate large prediction horizons NP in real time. NP 8Np 4NP PP CBF 1 8 4 4 4 2 64 16 14 10 3 512 64 33 26 4 4,096 256 48 26 5 32,768 1,024 67 26 6 262,144 4,096 81 68 -15- 16/26
    AV-3598 AT 7 2.097.152 16.384 116 68 8 16.777.216 65.536 178 68 9 134.217.728 262.144 237 178 10 1.073.741.824 1.048.576 313 178 11 8.589.934.592 4.194.304 489 178
    Table 2
    The method according to the invention makes it possible to take account of large prediction horizons NP for the controller 18 in the control unit 10, as a result of which the controller 18 is better able to react to transient processes and to be able to control such processes more quickly and with less overshoot. This is supported by the higher sample rate enabled.
    The inventive method has been described with reference to a multi-phase DC / DC converter 12 with three strands. It will be understood, however, that the method may be applied to a DC / DC converter 12 having fewer or more strings. -16- 17/26
    权利要求:
    Claims (6)
    [1]
    A method for model predictive control of a multiphase DC / DC converter (12) having a half-bridge with two switches (S1, S2, S3, S4, S5, S6) for each leg of the multi-phase DC / DC converter (12), wherein the switches (S1, S2, S3, S4, S5, S6) for generating a desired output variable (v2, i2) are controlled by a control unit (10), characterized in that the optimization problem of model predictive control in two Optimization problems are divided by implementing in the control unit (10) a model predictive output control and a model predictive inductor current control, for the output control, the strands of the multi-phase DC / DC converter (12) are combined into a single strand and from a time-discrete state space model is created and the output variable control based on a first cost function (Jv) of the optimization problem of the output control Input voltage (uv, k + i) of the next sampling step (k + 1) predicts for this single strand, the throttle current control is given as a default and the inductor current control from a second cost function (Jj) of the optimization problem of the inductor current control for the next sampling step (k +1) determines the necessary switch positions of the switches (S1, S2, S3, S4, S5, S6) of the strings of the multi-phase DC / DC converter (12).
    [2]
    2. The method according to claim 1, characterized in that a solution space in the form of a set of possible solutions of the optimization problem of the output control is examined in advance and solutions that can not occur are excreted from the solution space to reduce the solution space.
    [3]
    3. The method according to claim 2, characterized in that a simulation run with a predetermined control sequence of the output variable (v2, i2) is performed and the occurring input vectors (u,) of the DC / DC converter (12) are recorded and from the input vectors (u,) the input voltages (uv) of the output variable control are reconstructed and the occurring sequences of the input voltage (uv) are stored as a reduced solution space.
    [4]
    4. The method according to claim 1, characterized in that solutions of the optimization problem of the output variable control are represented as an integer linear combination of basic functions, wherein the basis functions for the first n sampling steps (k + n) have a width of one sampling step (k) and the basic functions for the following sampling steps (k + m), where m> n, have a width of an integer multiple of a sampling step (k). -17- 18/26 AV-3598 AT
    [5]
    5. The method according to claim 4, characterized in that the integer linear combinations are subject to the restriction that solutions from one sampling step (k) to the next sampling step (k + 1) must not switch more than one.
    [6]
    6. Model predictive controller of a multi-phase DC / DC converter (12) with a half bridge with two switches (S1, S2, S3, S4, S5, S6) for each phase of the multi-phase DC / DC converter (12) and with a control unit (10) which actuates the switches (S1, S2, S3, S4, S5, S6) for generating a desired output variable (v2, i2), characterized in that a model-predictive output variable regulator (15) is provided in the controller (18). and a model predictive inductor current regulator (16) are provided, wherein for the output variable regulator (15) for generating a discrete-time state space model, the strings of the multiphase DC / DC converter (12) are combined into a single string and the output variable regulator (15) based on a first Cost function (Jv) of the optimization problem of the output variable regulator (15) predicts the input voltage (uv, k + i) of the next sampling step (k + 1) for that single strand and the inductor current regulator (16) predicts from the predicted input 15 ngsspannung (uv, k + i) based on a second cost function (Jj) of the optimization problem of the inductor current regulator (16) for the next sampling step (k + 1) the necessary switch positions of the switches (S1, S2, S3, S4, S5, S6) Strings of the multi-phase DC / DC converter (12) determined and that the control unit (10) the DC / DC converter (12) in the next sampling step (k + 1), the determined switch positions. 20 -18- 19/26
    类似技术:
    公开号 | 公开日 | 专利标题
    AT513776B1|2015-09-15|Method and controller for model-predictive control of a multiphase DC / DC converter
    EP2534748B1|2018-10-17|Control of a modular multilevel converter with an observer for the currents and an estimator for the energy of the intermediate circuit
    AT15107U1|2017-01-15|Method and device for testing the drive train of at least partially electrically operated vehicles
    DE102008034109A1|2010-02-04|Circuit for simulating an electrical load
    DE102010007184A1|2010-11-25|PWM Rectifier
    DE102013212426A1|2014-12-31|Inverter arrangement with parallel-connected multi-stage converters and methods for their control
    Khaldi et al.2015|Fractional-order control of three level boost DC/DC converter used in hybrid energy storage system for electric vehicles
    DE102013203706B3|2014-06-05|Method of operating modular high frequency converter mounted in e.g. electric vehicle, involves replacing switching position model parameters that represent influence of switch positions of input-side half-bridge by related duty cycle
    EP3132515B1|2019-01-30|Converter for outputting reactive power, and method for controlling said converter
    DE102011078579A1|2013-01-10|Drive circuit and method for driving the input switching stage of a transformer circuit
    DE102017113468A1|2017-12-28|POWER CONVERTER WITH AN AUTOTRANSFORMER AND METHOD FOR CHANGING PERFORMANCE
    WO2018095659A1|2018-05-31|Regulation of an output current of a converter
    WO2021035269A1|2021-03-04|Method and controller for model predictive control of coupled converters
    DE102007060330A1|2009-06-25|Power supply for generating temporally specifiable, controlled and regulated current waveforms and method
    EP3138176B1|2021-05-05|Converter for symmetrical reactive power compensation, and a method for controlling same
    DE112018007312T5|2020-12-17|METHOD FOR PULSE WIDTH MODULATION OF A POWER CONVERTER AND A POWER CONVERTER
    AT520392B1|2020-08-15|Energy storage emulator and method for emulating an energy storage
    DE102015116271A1|2017-03-30|Method for operating a modular multilevel converter, modular multilevel converter and computer program
    DE102019204695A1|2020-10-08|COMPENSATION CIRCUIT FOR INTERFERENCE CURRENTS
    DE102018216749A1|2020-04-02|Method for controlling a series resonance converter
    DE102013019957A1|2015-06-11|Device for emulating the electrical behavior of electric three-phase machines
    DE102020214013A1|2021-05-27|POWER CONVERTER, CONTROL PROCEDURE AND CONTROL PROGRAM
    DE102009023626A1|2010-12-02|Device for compensation of harmonics
    DE102009008048A1|2010-08-19|Method for controlling a self-commutated mains converter of a voltage source inverter
    EP0491690B1|1994-06-08|Process for control of a multi-stage pulse converter
    同族专利:
    公开号 | 公开日
    CN106164689A|2016-11-23|
    CN106164689B|2019-04-19|
    JP6506775B2|2019-04-24|
    EP3129799A1|2017-02-15|
    AT513776B1|2015-09-15|
    HUE039394T2|2018-12-28|
    US10216153B2|2019-02-26|
    AT513776A3|2015-05-15|
    KR102267679B1|2021-06-22|
    US20170227937A1|2017-08-10|
    EP3129799B1|2018-07-25|
    PL3129799T3|2019-01-31|
    KR20160142388A|2016-12-12|
    WO2015154918A1|2015-10-15|
    JP2017516440A|2017-06-15|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    CN111538238A|2020-03-25|2020-08-14|西安电子科技大学|Method and system for switching sequence control and error analysis of first-order discrete switch|
    CN111781831A|2020-07-09|2020-10-16|济南大学|Dual-mode prediction control method and system of boost circuit based on state estimation|US4961048A|1989-08-03|1990-10-02|The United States Of America As Represented By The Secretary Of The Navy|MOSFET switching arrangement in a power converter unit|
    US5023542A|1989-08-03|1991-06-11|The United States Of America As Represented By The Secretary Of The Navy|Current limiting arrangement in a power converter|
    JPH04256102A|1991-02-08|1992-09-10|Toshiba Corp|Model estimation controller|
    US5754413A|1996-02-23|1998-05-19|Lucent Technologies Inc.|Reduced voltage stress asymmetrical DC-to-DC converter using first and second transformers having differing turns ratios|
    DE10259088B4|2002-12-17|2007-01-25|Infineon Technologies Ag|Resonance converter with voltage regulation and method for driving variable loads|
    JP4151014B2|2003-03-11|2008-09-17|株式会社デンソー|Isolated switching DC / DC converter|
    US7038432B2|2003-12-16|2006-05-02|Intersil Americas Inc.|Linear predictive controller|
    JP2006072733A|2004-09-02|2006-03-16|Fujitsu Ten Ltd|Model prediction controller and control method|
    JP4678215B2|2005-03-15|2011-04-27|サンケン電気株式会社|Switching power supply|
    WO2007059440A2|2005-11-11|2007-05-24|L & L Engineering Llc|Non-linear pwm controller for dc-to-dc converters|
    CN101305330A|2005-11-11|2008-11-12|L&L工程公司|Non-linear controller for switching power supply|
    US8598858B2|2007-05-29|2013-12-03|Nagasaki University, National University Corporation|Predictive control system|
    EP2528225B1|2010-01-22|2014-01-15|ABB Research Ltd.|Control of a rotating electric machine|
    EP2479994B1|2011-01-19|2017-03-15|BlackBerry Limited|Method and device for improved multi-layer data compression|
    US8553769B2|2011-01-19|2013-10-08|Blackberry Limited|Method and device for improved multi-layer data compression|
    US20140058787A1|2011-05-05|2014-02-27|Ron BANNER|Plan Choosing in Digital Commercial Print Workflows|
    US8503204B2|2011-08-05|2013-08-06|Infineon Technologies Ag|Power converter circuit|
    EP2621074A1|2012-01-24|2013-07-31|ABB Research Ltd.|Multicore implemented weight-function based predictive control of an electric converter|
    AT511270B1|2012-05-24|2015-07-15|Avl List Gmbh|Method and device for testing electric energy storage systems for driving vehicles|
    AT511520B1|2012-05-24|2017-06-15|Avl List Gmbh|Method and device for testing the drive train of at least partially electrically operated vehicles|
    WO2014006200A1|2012-07-06|2014-01-09|Abb Technology Ag|Controlling a modular converter|
    EP2891240A1|2012-08-28|2015-07-08|ABB Technology AG|Controlling a modular converter in two stages|
    CN103036460B|2012-11-26|2014-10-22|天津大学|Model prediction control method for three-level voltage-source-type converter|
    CN103117657B|2013-01-31|2015-06-03|浙江大学|Control method of full-bridge DC-DC system based on on-chip model predictive control|
    CN103595069B|2013-11-18|2015-07-29|河南师范大学|Photovoltaic generating system grid side converter model predictive control method under unbalance voltage|AT517652B1|2015-07-28|2018-04-15|Avl List Gmbh|Battery emulator and method for regulating the battery emulator|
    CN105391299B|2015-12-24|2019-02-01|西安理工大学|Buck converter pure strategy model predictive control method|
    EP3196714B1|2016-01-19|2020-12-16|dSPACE digital signal processing and control engineering GmbH|Simulation method and simulation device|
    CN106842958B|2017-03-20|2019-10-15|福州大学|A kind of distributed model predictive control method of back-to-back converter|
    CN107168319B|2017-06-01|2020-07-28|同济大学|Unmanned vehicle obstacle avoidance method based on model predictive control|
    CN107357168B|2017-06-01|2019-12-31|同济大学|Unmanned vehicle obstacle avoidance method based on opportunity constraint model predictive control|
    CN107546998B|2017-07-25|2019-12-10|华南理工大学|Switching type control method based on double-loop prediction control|
    AT520392B1|2017-09-04|2020-08-15|Avl List Gmbh|Energy storage emulator and method for emulating an energy storage|
    AT521666B1|2018-07-09|2022-02-15|Avl List Gmbh|Method and device for compensating for disturbance variables|
    EP3614218A1|2018-08-22|2020-02-26|Technische Hochschule Nuernberg Georg-Simon-Ohm|Cascaded continuous and finite model predictive control for mechatronic systems|
    CN109301823A|2018-10-31|2019-02-01|天津城建大学|A kind of electrical energy power quality disturbance compensation method based on Finite State Model predictive control strategy|
    EP3657659A1|2018-11-21|2020-05-27|Etel S. A..|Electrical power converter and method for controlling an electrical power converter|
    US10700605B1|2018-12-12|2020-06-30|Infineon Technologies Austria Ag|Electrical power converter with predictor|
    US20200333403A1|2019-04-20|2020-10-22|University Of Tennessee Research Foundation|Power electronics based reconfigurable load tester|
    CN110244567B|2019-07-04|2021-09-24|武汉大学|Rapid model prediction control method based on extended instantaneous reactive power theory|
    AT522895A1|2019-08-30|2021-03-15|Avl List Gmbh|Method and controller for model predictive control of a converter|
    AT523984A1|2020-07-02|2022-01-15|Avl List Gmbh|Model predictive control of a power converter|
    CN112803804A|2021-02-01|2021-05-14|中南大学|Single-phase three-level rectifier parameter compensation temperature balance control method and system|
    法律状态:
    优先权:
    申请号 | 申请日 | 专利标题
    ATA50265/2014A|AT513776B1|2014-04-08|2014-04-08|Method and controller for model-predictive control of a multiphase DC / DC converter|ATA50265/2014A| AT513776B1|2014-04-08|2014-04-08|Method and controller for model-predictive control of a multiphase DC / DC converter|
    US15/302,051| US10216153B2|2014-04-08|2015-03-03|Method and controller for model predictive control of a multi-phase DC/DC converter|
    HUE15707640A| HUE039394T2|2014-04-08|2015-03-03|Method and controller for the model-predictive control of a multi-phase dc/dc-converter|
    JP2016559923A| JP6506775B2|2014-04-08|2015-03-03|Method and controller for model predictive control of multiphase DC / DC converter|
    CN201580016957.8A| CN106164689B|2014-04-08|2015-03-03|The method and controller of Model Predictive Control for multiphase DC/DC converter|
    PCT/EP2015/054415| WO2015154918A1|2014-04-08|2015-03-03|Method and controller for model predictive control of a multi-phase dc/dc converter|
    EP15707640.7A| EP3129799B1|2014-04-08|2015-03-03|Method and controller for the model-predictive control of a multi-phase dc/dc-converter|
    KR1020167031129A| KR102267679B1|2014-04-08|2015-03-03|Method and controller for model predictive control of a multi-phase dc/dc converter|
    PL15707640T| PL3129799T3|2014-04-08|2015-03-03|Method and controller for the model-predictive control of a multi-phase dc/dc-converter|
    [返回顶部]