WO2021210274A1

WO2021210274A1 - Power conversion system and method for controlling power conversion system

Info

Publication number: WO2021210274A1
Application number: PCT/JP2021/006994
Authority: WO
Inventors: 隆一小川; 昌司滝口
Original assignee: 株式会社明電舎
Priority date: 2020-04-16
Filing date: 2021-02-25
Publication date: 2021-10-21
Also published as: JP2021170882A; JP6885489B1

Abstract

In a prediction unit (7), a plurality of tentative voltage vectors are set for each output cycle during a prediction period, and predicted currents for the tentative voltage vectors are calculated on the basis of a correction current value obtained by adding a current variation amount (Δidc, Δiqc) to the detected current during a calculation cycle. The prediction unit (7): calculates the tentative voltage vectors and an evaluation function (J) for the predicted current; determines a combination of tentative voltage vectors for which the evaluation function (J) is at a maximum level from among the combinations of the tentative voltage vectors of the prediction period; and outputs, as a command voltage vector (V^*), the tentative voltage vector to actually be outputted, from among the combinations of tentative voltage vectors for which the evaluation function (J) is at the maximum level. A gate signal determination unit (8) outputs a gate signal (g) for outputting the voltage represented by the command voltage vector (V^*) from a power converter (3). A control having high current control performance and low switching frequency is thereby performed in a power conversion system using a model prediction control (MPC).

Description

Power conversion system and control method of power conversion system

The present invention relates to a current control method in a power conversion system that outputs a voltage using a power converter.

Consider a system in which the input three-phase AC voltage is converted into a DC voltage by a rectifier (AC-DC converter), and the DC voltage is output as an AC voltage with a desired frequency and amplitude by an inverter (power converter). The inverter controls the output voltage by switching the semiconductor element. At this time, the voltage that can be output by the inverter becomes a discrete value.

In such a system, current control may be used to obtain the desired value of the output current. In a general configuration of current control, first, an ACR (Auto Current Regulator) determines a command voltage with a continuous value. However, continuous voltage values cannot actually be output. Therefore, a triangular wave comparison PWM (Pulse Width Modulation) is performed to determine a discrete output voltage that can achieve the command voltage on average.

However, there are current control methods other than the configuration using ACR and triangular wave comparison PWM. An example of this is the MPC (Model Predictive Control: model prediction control) that is the subject of the present application. The MPC predicts the change in current with respect to each assumed output voltage, and adopts a voltage as the actual output so that the locus of the output current follows the command current most. The feature of MPC is that the discrete output voltage is considered in advance and the instantaneous current can be predicted.

Regarding MPC, Non-Patent Document 1 and Patent Document 1 are disclosed. Non-Patent Document 1 describes a basic control configuration of MPC, a method of predicting current using a motor state equation, and a method of evaluating predicted current and voltage vector. Regarding the evaluation method, by considering items such as the deviation between the command current and each predicted current, the maximum value of the total deviation, and the number of changes in the voltage vector, an output current close to the command current can be obtained and the number of voltage changes. The output voltage with a small (number of switchings) can be determined.

Patent Document 1 describes an MPC control method for performing current prediction calculation up to i (n + 2). In Patent Document 1, the detection time is set to n, and the current change during the calculation time is predicted to obtain i (n + 1). After that, i (n + 2) is calculated from i (n + 1) in consideration of the voltage vector branch of the inverter, and the voltage vector is determined by the evaluation of i (n + 2).

Non-Patent Document 1 does not consider the change in current during the calculation time.

In Patent Document 1, the detected current is i (n), and the current change during the calculation time is predicted from the detected current to calculate the current i (n + 1). Then, i (n + 2) is calculated using i (n + 1). In this way, the predicted current of the previous current is obtained. However, it does not mention a general and specific calculation method such as a case where a voltage vector for a plurality of cycles is determined by one calculation when predicting n + 2 or later.

From the above, it is an issue to control high current control performance and low switching frequency in a power conversion system using model predictive control (MPC).

JP-A-2010-252433

The present invention has been devised in view of the above-mentioned conventional problems, and one aspect thereof is an upper control unit that generates a command current based on a command value and each output cycle during the prediction period by model prediction control. A plurality of assumed voltage vectors are set in, the predicted current of the assumed voltage vector is calculated based on the correction current value obtained by adding the amount of change in the current during the calculation cycle to the detected current, and the assumed voltage vector and the predicted current are evaluated. The function is calculated, the combination of the assumed voltage vectors whose evaluation function is the highest is determined from the combination of the assumed voltage vectors in the prediction period, and the combination of the assumed voltage vectors whose evaluation function is the highest is determined. A predictor that outputs the assumed voltage vector that is actually output from, as a command voltage vector, a gate signal determination unit that outputs a gate signal for outputting the voltage expressed by the command voltage vector from the power converter, and a gate signal. It is characterized by being equipped with a power converter that is driven and controlled based on the above.

Further, as one aspect thereof, the amount of change in current during the calculation cycle is calculated by the following equation (5).

Δidc: Current change amount during calculation cycle (d-axis)
Δiqc: Current change amount during calculation cycle (q-axis)
Tc: Output cycle (= calculation cycle)
id: Detected d-axis current iq: Detected q-axis current vdz: d-axis voltage vector Previous value vqz: q-axis voltage vector Previous value R: Winding resistance Ld: d-axis inductance Lq: q-axis inductance ωr: Detected electrical angular velocity ψ: Permanent magnet interlinkage magnetic flux number.

Further, as another aspect, a plurality of output cycles during the prediction period are provided by a host control unit that generates a command current based on the command value and model prediction control that determines voltage vectors of a plurality of output cycles in one calculation cycle. The assumed voltage vector is set, the predicted current of the assumed voltage vector is calculated based on the correction current value obtained by adding the amount of change in the current during the calculation cycle to the detected current, and the evaluation function of the assumed voltage vector and the predicted current is calculated. Then, from the prediction unit that outputs the combination of the assumed voltage vectors having the highest evaluation function as the command voltage vector matrix from the combinations of the assumed voltage vectors in the prediction period, and the command voltage vector matrix having a plurality of output cycles. This time, a read unit that selects the voltage vector of the output cycle and outputs it as an output voltage vector, a gate signal determination unit that outputs a gate signal for outputting the voltage expressed by the output voltage vector from the power converter, and a gate. It is characterized by being equipped with a power converter that is driven and controlled based on a signal.

Further, as one aspect thereof, the current change amount during the calculation cycle is calculated by calculating the current change amount during each output cycle by the following equation (21), and adding the current change amounts of all the output cycles during the calculation cycle. It is characterized in that the value is set to the value.

Δidc: Current change amount during output cycle (d-axis)
Δiqc: Current change amount during output cycle (q-axis)
Tc: Output cycle id': Detected d-axis current or d-axis current iq': Detected q-axis current or q-axis current vdz: d-axis voltage vector Previous value vqz: q-axis voltage vector Previous value R: Winding Resistance Ld: d-axis inductance Lq: q-axis inductance ωr: Detected electric angular velocity ψ: Permanent magnet interlinkage magnetic flux number.

Further, as one aspect thereof, the correction current value is characterized by adding the amount of change in the current during the calculation cycle and the amount of change in the current in the timing shift period of the output voltage change with respect to the current detection to the detected current. do.

Further, as one aspect thereof, the amount of current change in the timing shift period of the output voltage change with respect to the current detection is calculated by the following equation (17).

Δiddel: Current change amount (d-axis) during the timing shift period of the output voltage change with respect to the current detection
Δiqdel: Current change amount (q-axis) during the timing shift period of the output voltage change with respect to the current detection
Tdel: Timing shift period of output voltage change with respect to current detection id: Detected d-axis current iq: Detected q-axis current vdzz: d-axis voltage vector Pre-previous value vqz: q-axis voltage vector Pre-previous value R: Winding resistance Ld: d-axis inductance Lq: q-axis inductance ωr: detected electric angular velocity ψ: permanent magnet interlinkage magnetic flux number.

According to the present invention, in a power conversion system using model predictive control (MPC), it is possible to control high current control performance and low switching frequency.

The block diagram of the power conversion system in Embodiments 1 to 4. The figure which shows the relationship between the voltage vector of a two-level inverter and a phase voltage. The figure which shows the relationship between the detected current and the predicted current. The block diagram which shows the MPC in Embodiment 1. FIG. The flowchart which shows the processing of the prediction part in Embodiment 1. The block diagram which shows the MPC in Embodiment 2. The flowchart which shows the action of the prediction part in Embodiment 2. The block diagram which shows MPC in Embodiment 3. The flowchart which shows the processing of the prediction part in Embodiment 3. The flowchart which shows the processing of the reading part in Embodiment 3. The block diagram which shows the MPC in Embodiment 4. The flowchart which shows the processing of the prediction part in Embodiment 4.

Hereinafter, embodiments 1 to 4 of the power conversion system using the model predictive control (MPC) in the present invention will be described in detail with reference to FIGS. 1 to 12.

[Embodiment 1]
FIG. 1 shows a configuration diagram of the power conversion system according to the first embodiment. The upper control unit 1 indicates a control existing upstream of the MPC 2. For example, a speed control calculation is performed using a speed command and a detection speed based on information such as the operation amount of the panel / panel and the accelerator opening, and the command is given. Generates current i ^*.

^{The command current i *} , the detection current i, and the detection phase θ output from the upper control unit 1 are input to the MPC (model prediction control unit) 2. However, the detection phase θ is an electric angle. For example, when the detected value is the mechanical position of the motor, the conversion is appropriately performed using the number of pole pairs.

MPC2 performs an operation to predict the current from the next time onward based on the model parameters, and determines the voltage vector based on the result. A gate signal g for outputting according to the determined voltage vector is output from the MPC 2, and the power converter (for example, an inverter, hereinafter referred to as an inverter) 3 is driven by the gate signal g.

The inverter 3 is connected to a load 4 such as a motor, and a voltage corresponding to the gate signal g is applied to the load 4.

FIG. 1 is a typical system configuration example of a power conversion system using MPC, and the application target of the present invention is not limited to this. For example, a DC / AC converter that regenerates a power source may have a configuration in which current control is performed based on model prediction, a configuration in which an electric angle is estimated by a PLL (Phase Locked Loop) and used in an MPC, or a single-phase configuration may be used. What is important is that the voltage vector is determined based on the current prediction result by the model prediction control, and the power converter is driven by the voltage vector.

Definitions of cycles, periods, etc. related to control in this specification are described below.
Predicted step: Predicted step time of current (time width from time i (n) to time i (n + 1))
Prediction period: Time width of the earliest time for predicting current and time of i (n) Output cycle: Cycle for switching voltage output Calculation cycle: Cycle for performing prediction calculation.

Since it is selected whether to switch for each output cycle, the prediction step and the output cycle are treated as the same time width below. Further, in the first embodiment, the calculation cycle and the output cycle are the same.

In MPC2, the voltage vector of the inverter 3 is assumed, the predicted current when the assumed voltage vector is output is obtained, and the voltage vector that gives the optimum prediction result is adopted and used as the actual output.

Since MPC2 evaluates the predicted current to determine the output voltage, the current prediction accuracy is important. The current prediction operation is based on the equation of state. For example, when PMSM (Permanent Magnet Synchronous Motor) is used for the load, the equation of state in Eq. (1) is obtained. Here, Ld is the d-axis inductance, Lq is the q-axis inductance, R is the winding resistance, ψ is the number of permanent magnet interlinkage magnetic fluxes, and ωr is the detected electric angular velocity. In the following, the equation (1), that is, PMSM will be used as a reference, but from the essence of the present invention, the present invention may be applied to other than the case where the equation of state is the equation (1).

Equation (1) is a continuous equation of state, but if it is made into a discrete equation of state in order to handle it in a discrete control system, it becomes equation (2). Equation I in Eq. (2) represents an identity matrix. Alternatively, the product of the slope and the output period Tc may be simply taken and treated as an approximate amount of change, and the equation (3) may be used.

From equations (2) and (3), the current id (n), iq (n) and voltage corresponding to the current time are used to predict the current id (n + 1) and iq (n + 1) one cycle ahead of the output cycle. vd (n) and vq (n) are required.

For the voltages vd (n) and vq (n), the voltage vectors assumed in the previous stage of the prediction calculation are used. For example, in a two-level inverter with a DC voltage vdc, the voltage vector is shown in FIG. 2, so that the assumed voltage vector can be converted into a voltage on the dq axis by the calculation of Eq. (4). Here, θ (n) is the motor phase. The phase may be advanced and compensated according to the angular frequency, but detailed calculation will be described later.

On the other hand, the detected values are used as a reference for the currents id (n) and iq (n). The n + a-th predicted value can be used for the calculation of obtaining the current id (n + a + 1) and iq (n + a + 1) from the current id (n + a) and iq (n + a) in the middle of prediction, but the n-th predicted value is the prediction start point. The currents id (n) and iq (n) cannot be assumed and must be based on the detected value. Although there may be a configuration in which an estimated value is used instead of using the detected value-based currents id (n) and iq (n), in practice, correction using the detected value or a detected value-based current is applied every multiple cycles. Requires the configuration to be used. That is, in any case, the detected value is used as a reference.

However, although the values based on the detected values are used for the currents id (n) and iq (n), if the detected values themselves are used, the prediction accuracy will be lowered.

Figure 3 shows the relationship between the detected current and the predicted current. The solid line shows the output current, and the dotted line shows the predicted current. Tc is the output cycle. The predicted current i (n) is the d-axis and q-axis currents, but the notation of d and q is omitted because it shows the property of being applicable to both the d-axis and q-axis currents.

In the second cycle, the current after the third cycle is predicted based on the voltage vector assumed to be the current detected at the end of the first cycle. Moreover, since there are a plurality of types of voltage vectors to be assumed, the predicted current of the dotted line is branched into a plurality of types.

FIG. 3A shows a case where the detected value is used for i (n). Looking at FIG. 3, there is a time difference of the output cycle Tc between the detection time and the time of i (n). This time difference is the calculation time from the detected value to the determination of the output voltage from the next time onward, and it is impossible to make the detection time and the time of i (n) the same time.

However, during the calculation time, the voltage vector determined in the previous output cycle is output, and the current changes from the detected value from the detection time to the time of i (n). Therefore, if the detected value is used for i (n), the value of i (n) used in the control and the value of the actual output current at the time of i (n) will be different. As a result, there is a concern that the current prediction accuracy will decrease. If the current prediction accuracy is lowered, the validity of the subsequent current evaluation is also lowered, which causes an increase in current ripple and an increase in switching frequency.

FIG. 3B shows a case where the value obtained by adding the amount of change in current during the calculation time to the detected value is used as i (n). The current prediction calculation from the detection time to the time i (n) is performed using the voltage vector determined in the previous cycle, and the current change during the calculation time is predicted. The broken line shows the change in current. By doing this, accurate i (n) is obtained, and the current prediction accuracy is improved as compared with FIG. 3 (a).

As described above, by predicting the current change during the calculation time, the current prediction accuracy can be improved, and effects such as reduction of current ripple and reduction of switching frequency can be obtained.

In the following, the configuration for predicting the current change during the calculation time will be described. FIG. 4 shows a configuration diagram of MPC2 according to the first embodiment. For convenience, the command current i ^* is divided into a command d-axis current id ^* and a command q-axis current iq ^*. The three-phase two-phase conversion unit 5 converts the three-phase detection current i into UVW / dq based on the detection phase θ, and converts it into the detection d-axis current id and the detection q-axis current iq.

The differentiator 6 outputs the detected electric angular velocity ωr by differentiating the detection phase θ or pseudo-differentiating processing.

The command d-axis current id ^* , command q-axis current iq ^* , detected d-axis current id, detected q-axis current iq, detected phase θ, detected electric angular velocity ωr, and command voltage vector previous value Vz ^* are input to the prediction unit 7. .. The prediction unit 7 performs a current prediction calculation and evaluates the prediction result, and outputs a command voltage vector V ^* which is an output voltage vector from the next time onward. The command voltage vector V ^* is input to the prediction unit 7 as the command voltage vector previous value Vz ^{* after one delay, and is used for the next prediction calculation.}

The gate signal determination unit 8 determines the gate signal g for outputting the voltage represented by the ^{command voltage vector V *} from the inverter based on the circuit configuration of the system, and inserts the dead time. Then, the gate signal g becomes the output of MPC2. In the first embodiment, the configuration of the gate signal determination unit 8 is not particularly limited as long as the voltage according to the ^{command voltage vector V * is output.}

What is important in FIG. 4 is that the prediction unit 7 dynamically performs the current prediction calculation, and the configuration is not limited to FIG. For example, a configuration in which the estimated electric angular velocity is used without using the detected electric angular velocity ωr is also allowed.

FIG. 5 is a flowchart showing the processing of the prediction unit 7. The prediction unit 7 performs the following processing.
(A) Predict the change in current during the calculation time.
(B) Assuming the voltage vector of the output cycle from the next time onward, the predicted current in that case is obtained.
(C) Evaluate the voltage vector assumed to be the predicted current.
(D) A voltage vector with a good evaluation result is adopted and used as an output.
The process (A) is the essence of the present invention, and the processes (B), (C), and (D) are basic processes in the MPC.

In 1-1, the command d-axis current id ^* , the command q-axis current iq ^* , the detected d-axis current id, the detected q-axis current iq, the detected phase θ, the detected electric angular velocity ωr, and the command voltage vector previous value Vz ^* are input. ..

In 1-2, the calculation time is obtained from the command voltage vector previous value Vz ^* , the detected d-axis current id, and the detected q-axis current iq, and here, the current changes Δidc and Δiqc during the output cycle Tc are obtained. This operation may be performed as in Eq. (5) based on Eq. (3). However, the matrices A and B and the vector eψ are the same as in Eq. (1), and vdz and vqz are values obtained by converting the command voltage vector previous value Vz ^* into the dq axis voltage.

In 1-3, the current change amounts Δidc and Δiqc during the calculation time obtained in 1-2 are added to the detected d-axis current id and the detected q-axis current iq, and these are added to the id (n), which is the starting point of the prediction calculation. Let it be iq (n). Let these id (n) and iq (n) be the correction current values.

1-2 and 1-3 correspond to the above (A), and the current change from the detection time to the id (n) and iq (n) times is taken into consideration by these processes, and the accuracy of the current prediction calculation is improved. do.

1-4 sets the assumed voltage vector V'(n + a) for each output cycle during the prediction period. a is a counter value indicating which output cycle is considered from the start of prediction, and increases from 0 to N-1 as the calculation progresses. N is the number of output cycles during the prediction period. In the first embodiment, the setting method is not particularly limited. That is, it may be a voltage vector that changes with little switching according to the output up to the previous time, or one may be selected from all the voltage vectors that can be output.

1-5 is a prediction of the amount of current change when the assumed voltage vector is output. From the hypothetical voltage vectors V'(n + a), id (n + a), and iq (n + a), the current change amounts Δid and Δiq during the output period Tc are obtained. id (n + a) and iq (n + a) are correction current values when a is 0, and predicted currents when a is 1 or more. Similar to 1-2, the calculation may be as shown in Eq. (6).

Here, vd'(n + a) and vq'(n + a) are obtained by converting the assumed voltage vector v'(n + a) into a voltage on the dq axis. However, although the motor phase is required to obtain the dq-axis voltage, the phase lead compensation may be performed for the motor phase according to the angular velocity. In this case, the calculation is in Eq. (7).

θ is the detection phase, ωr is the detected electrical angular velocity, and Tc is the output period. Further, 1 + a + Kθ are 1 indicating the calculation time (= output cycle), a indicating how many cycles ahead from the nth cycle, and a phase for using the phase at an arbitrary time in the output cycle as a voltage reference. The adjustment term Kθ and the adjustment term Kθ are added. Kθ is determined by Eq. (8). When Kθ = Tc / 2, the time in the middle of the output cycle becomes the reference for the dq coordinate conversion.

In 1-6, the current change amounts Δid and Δiq during the calculation time obtained in 1-5 are added to id (n + a) and iq (n + a), and the predicted current id (1 cycle ahead, that is, the output cycle Tc ahead) ( n + a + 1) and iq (n + a + 1) are calculated. If there is a previous branch, this current is used to obtain the next n + a + second predicted current.

1-4 to 1-6 correspond to the above process (B).

1-7 evaluates the assumed voltage vector V'(n + a), the predicted current id (n + a + 1), and iq (n + a + 1). The first embodiment can be applied to any evaluation method, and may be evaluated as follows, for example.

The evaluation function J is calculated using the assumed voltage vector V'(n + a), the predicted current id (n + a + 1), and the iq (n + a + 1). The evaluation function J is a loss function that becomes a large value when it is an undesired prediction, and is calculated by weighting a plurality of evaluation criteria. Equations (9) and (10) on condition that the deviation between the predicted current and the command current is small, equations (11) and (12) for preventing the deviation between the predicted current and the command current from exceeding a certain level, switching. Eq. (13), which gives a penalty to the voltage vector change so that the number of times does not increase, is each evaluation standard. However, idmax and iqmax are the maximum tolerances of the d-axis current and the q-axis current, respectively. Finally, the evaluation function J is obtained by the equation (14) using the weighting coefficients Wid, Wiq, Widmax, Wiqmax, and Wv.

In the actual 1-7 calculation, the evaluation function values up to the previous time are retained, and only the calculations related to the assumed voltage vector V'(n + a), predicted current id (n + a + 1), and iq (n + a + 1) for this time are performed. It may be weighted and added.

The evaluation function described above is an example, and is not limited to the calculation method of the first embodiment. Even with the same evaluation content, for example, Qd and Qq in Eqs. (11) and (12) may not be set to two values of 0 and 1, but continuous values may be given so that the larger the deviation from the command, the larger the value. .. Further, in the case of PMSM, from the derivation of the motor torque T of the equation (15), the equation (16) may be used as an evaluation standard instead of the equation (9). However, Pn in Eq. (15) is the pole logarithm, and T ^{* in} Eq. (16) is the command torque.

In the above evaluation criteria, not only the error between the predicted current and the command current is small, but also the evaluation function becomes small for the voltage vector in which the number of times the voltage vector is changed, that is, the number of times of switching is small. Therefore, according to this evaluation standard, current control can be performed while reducing the switching frequency.

Also, the control performance can be changed by adjusting the weighting factor. For example, if Wv is made smaller, the current control performance can be improved instead of increasing the switching frequency, and if Wiq is made larger than Wid, the ripple of iq can be made smaller instead of increasing the ripple of id. The process of 1-7 corresponds to the above (C).

1-8 is a branching process of whether or not the calculation is performed assuming all the voltage vectors. When the number of output cycles during the prediction period is N, the voltage vector to be considered is up to V'(n + N-1), so if prediction calculation and evaluation are performed for all V'(n + N-1) to be considered, all branches are considered. If so, proceed to 1-9. If there is still a voltage vector to consider, go back to 1-4 and assume the next voltage vector.

In 1-9, the evaluation results are compared to determine the best combination of voltage vectors. When the evaluation function J is defined as in Eq. (14), V (n) to V (n + N-1), which is the minimum J, is the best combination of voltage vectors V (n) to V (n + N-1). ).

In 1-10, the voltage actually output from the best combination of voltage vectors V (n) to V (n + N-1) is substituted ^{into the command voltage vector V *.} Here, the number of output cycles during the calculation time is set to 1, and in 1-10, only V (n) is substituted.
1-9 and 1-10 correspond to the above-mentioned process (D).

In 1-11, the command voltage vector V ^* is output. The above is the operation of the flowchart of FIG.

FIG. 5 is an example of voltage vector determination based on model prediction. The detailed calculation procedure of the prediction unit 7 of the first embodiment is not limited to the flowchart of FIG. What is important is that the current values id (n) and iq (n), which are the starting points of the predicted current calculation / evaluation, are not the detected values, but the current values obtained by adding the current change amount during the calculation cycle to the detected values.

As shown above, according to the first embodiment, control that achieves high current control performance and low switching frequency by performing MPC that predicts the current change during the calculation time based on FIGS. 4 and 5. It can be performed.

In addition, there is an advantage that the forecast period can be applied to the prior literature when the prediction period is general.

[Embodiment 2]
In the first embodiment, the current id (n) and iq (n), which are the starting points of the prediction calculation, are set in consideration of the current change during the calculation time, and the current prediction accuracy is improved.

However, in reality, there is a time lag such as communication between the determination of the gate signal g and the actual output of the voltage from the inverter 3.

Further, since high-speed current vibration (ringing) occurs at the moment of switching of the inverter 3, there is a possibility that the current in the middle of vibration is detected if the detection is performed by the interrupt of the output cycle which is the switching of the voltage vector. As a countermeasure, the current detection interrupt may be set at a timing shifted from the output cycle interrupt.

For the above reasons, the timing of the detected current and the predicted current may be different. Although this deviation is due to a plurality of reasons as described above, in the present specification, it is collectively referred to as a delay time Tdel. The delay in this case is the delay of the predicted current change with respect to the detection timing, but the predicted current change timing can be regarded as the change timing of the output voltage. That is, the timing shift period of the output voltage change with respect to the current detection is the delay time Tdel.

In MPC2, which is required to highly predict the instantaneous current, the current prediction accuracy is lowered even if the delay time Tdel is about several μs. Therefore, in the second embodiment, the current change during this time is taken into consideration.

FIG. 6 shows a configuration diagram of MPC2 according to the second embodiment. Since the voltage determined two times before is output instead of the voltage determined last time during the delay time Tdel, the command voltage vector two times before the previous value Vzz ^* is added to the input of the prediction unit 7. Other configurations are the same as those in the first embodiment.

It should be noted that, if it is understood that the voltage vector and the previous voltage vector of the last but one, even without a command voltage vector the second preceding value Vzz ^* and the command voltage vector previous value Vz ^*, the command voltage vector the second preceding value Vzz ^* and the command voltage vector before last value Vzz Different inputs may be used, such as a pattern of change ^{from *} to the command voltage vector previous value Vz ^*.

FIG. 7 shows a flowchart of the prediction unit 7 in the second embodiment. Current obtained by adding not only the amount of current change during the calculation cycle but also the amount of current change during the timing shift period (delay time Tdel) of the output voltage change with respect to current detection to id (n) and iq (n), which are the starting points of the prediction calculation. The process is the same as in FIG. 5 except that the value is used.

That is, two new processes 2-2 and 2-3 have been added between 1-1 and 1-2 in FIG. However, 2-4 and 2-5 are partially changed according to the added 2-2 and 2-3. The added process is the prediction of the amount of current change during the delay time Tdel.

In 2-2, the current change amounts Δiddel and Δiqdel in the delay time Tdel are calculated from the command voltage vector pre-previous value Vzz ^* , the detected d-axis current id, and the detected q-axis current iq. Based on the equation (3), the calculation may be performed as in the equation (17). However, vdzz and vqzz are conversion values of the command voltage vector pre-previous value Vzz ^{* to} the dq-axis voltage.

In 2-3, the current change amounts Δiddel and Δiqdel are added to the detected d-axis current id and the detected q-axis current iq to obtain the current values id'and iq'.

Since the timing of detection and output was aligned by the correction of 2-2, 2-3, the current values Δidc, Δiqc during the calculation time were used in 2-4 and 2-5 using these current values id'and iq'. And the start points id (n) and iq (n) of the prediction are calculated.

From 2-4 onward, the assumption of the voltage vector, the current prediction calculation / evaluation, and the adoption of the best voltage vector are performed in the same manner as in FIG. 5 of the first embodiment. The above is the operation of the flowchart of FIG. That is, in the second embodiment, the value obtained by adding the current change amounts Δidc and Δiqc during the calculation cycle and the current change amounts Δiddel and Δiqdel during the timing shift period of the output voltage change with respect to the current detection as the correction current value is used as the correction current value. There is.

By the way, the following corrections may be made to the phase lead correction. The phase lead for the delay time Tdel is taken into consideration by using Eq. (18) for the correction motor phase θ'used for 2-4 and 2-7 and Eq. (19) for the motor phase θ (n + a) used for 2-7. be able to.

As shown above, according to the second embodiment, the MPC is performed based on FIGS. 6 and 7 in which the current change during the calculation time is predicted and the timing difference between the current detection and the output voltage change is taken into consideration. Therefore, it is possible to perform control that achieves high current control performance, low switching frequency, and current detection at a timing that is not affected by ringing.

In addition, the prior literature has the advantage of considering the difference between the detection time and the output voltage change time, which is applicable when the prediction period is general.

[Embodiment 3]
FIG. 8 shows a configuration diagram of MPC2 according to the third embodiment. In the third embodiment, the prediction unit 7 determines a voltage vector having a plurality of output cycles and uses it as a command. ^{Therefore, unlike the first embodiment, V *,} which is the output of the prediction unit 7, is not a single voltage vector but a command voltage vector matrix. Similarly, the previous value Vz ^* is also the previous value of the command voltage vector matrix.

Further, a reading unit 9 is provided between the prediction unit 7 and the gate signal determination unit 8. The reading unit 9 selects the voltage vector of the output cycle this time from the command voltage vector matrix V ^* for a plurality of output cycles, and determines the ^{output voltage vector V **.} The input of the reading unit 9 is the command voltage vector matrix V ^* and the counter previous value Cz, and the output is the output voltage vector V ^** and the counter value C.

It should be noted here that the prediction unit 7 operates in the calculation cycle and the reading unit 9 operates in the output cycle. The one-time delay block also means one-time delay in each cycle of the input destination.

An important point of the third embodiment is that a mechanism for determining and reading out voltage vectors for a plurality of cycles is provided for the first embodiment, and the detailed mounting configuration is not limited to FIG.

The first and second embodiments describe a current prediction calculation method when the calculation cycle and the output cycle are equal, that is, when the output voltage of the inverter 3 does not change during the calculation time.

However, in MPC2, shortening the output cycle is expected to improve the current control performance, and considering the calculation load of the current prediction calculation, shortening the calculation cycle may make it impossible to implement. It is desirable to be able to support configurations that determine voltage vectors with multiple output cycles.

In the third embodiment, a configuration for predicting a current change during the calculation time when the number of output cycles during the calculation cycle is Ncal is examined. Since Ncalc is set to the number of output cycles N or less during the prediction period, there is a relationship of equation (20).

In handling the voltage vectors for a plurality of cycles, the output of the prediction unit 7 is used as the matrix information of the plurality of voltage vectors as shown in FIG. It is necessary to add a mechanism to read one. In addition, consideration of the current change during the calculation time, which is the essence of the present invention, must be taken for the voltage vector for a plurality of cycles.

FIG. 9 shows a flowchart of the prediction unit 7 in the third embodiment. Compared with FIG. 5 of the first embodiment, the output of the prediction unit 7 is a voltage vector for a plurality of output cycles, and the method of predicting the current change during the calculation time is also different accordingly. FIG. 9 operates at the calculation cycle Tcalc.

3-1 is the same as 1-1 in FIG. However, the command voltage vector previous value Vz ^* is changed to the command voltage vector matrix previous value Vz ^*.

In 3-2, determine the initial value of the repetitive calculation considering the current change during the calculation time. The variable b that manages the repetition is 1, and the temporary values id'and iq'of the dq-axis current are the detected d-axis flow id and the detected q-axis current iq as initial values.

In 3-3, the b-th of the command voltage vector matrix previous value Vz ^{* is Vz'.} Here, the command voltage vector matrix previous value Vz ^* has voltage vectors from the first to the Ncalc th, and the reading unit 9 outputs the command voltage vector matrix previous value Vz ^* in order from the first.

In 3-4, the current change amounts Δidc and Δiqc during the output cycle Tc are obtained from Vz', id'and iq'. Based on the equation (3), the calculation may be performed as in the equation (21). However, vdz'and vqz' are converted values of Vz'to the dq axis voltage. id'iq'is the detected d-axis current and detected q-axis current when b = 1, and the d-axis current and q-axis current calculated in 3-5 in the previous cycle when b = 2 to Ncalc. ..

In 3-5, Δidc and Δiqc, which are the amounts of current changes during the output cycle Tc, are added to id'and iq'.

3-6 is a branch of whether or not all output cycles during the calculation time are taken into consideration. If b has reached the number of output cycles Ncalc in the calculation cycle, it means that all the current changes in the calculation cycle have been added, and the process proceeds to 3-8. If b is less than the number of output cycles Ncalc, b is incremented by 1 at 3-7 to return to 3-3, and the current change amounts Δidc and Δiqc of the next output cycle are calculated. That is, when b = 1 to Ncalc, the current change amounts Δidc and Δiqc in each output cycle Tc are calculated, and the value obtained by adding the current change amounts in all the output cycles in the calculation cycle is the current change amount in the calculation cycle. Become. Further, id'iq'calculated in 3-5 when b = Ncalc is the correction current value.

In 3-8, the correction current values id'and iq', which are obtained by adding the current change amounts Δidc and Δiqc for all the output cycles during the calculation time, are substituted into the currents id (n) and iq (n) which are the starting points of the prediction calculation.

3-9 to 3-14 are the same processes as 1-4 to 1-9 of the first embodiment, and determine the assumption of the voltage vector and the prediction / evaluation of the current.

3-15 is modified from the first embodiment. Instead of substituting only the first V (n) from the adopted voltage vector, the voltage vectors up to the Ncalth number of output cycles during the calculation time are substituted into the command voltage vector matrix V ^* . The above is the operation of the flowchart of FIG.

FIG. 10 shows a flowchart of the reading unit 9 in the third embodiment. In FIG. 10, a counter is used to determine the ^{output voltage vector V **} , which is the voltage vector of the current output cycle, from the ^{command voltage vector matrix V * having voltage vector information for a plurality of cycles.} The configuration is not limited to FIG. 10 as long as the voltage vector of the current output cycle can be read from the voltage vectors of a plurality of cycles.

The flowchart of the reading unit 9 of FIG. 10 will be described. FIG. 10 operates with an output cycle Tc.

In R-1, the command voltage vector matrix V ^* and the counter previous value Cz are input. In R-2, the counter previous value Cz is substituted for the counter value C. In R-3, ^{the voltage vector V *} [C] of the counter value C (Cth) in the ^{command voltage vector matrix V *} is substituted into the output voltage vector V ^**.

R-4 is a branch process based on the counter value C. If the counter value C has reached the number of output cycles Ncalc in the calculation cycle, the counter value C is returned to 1 by R-5. If the counter value C is not Ncalc, the counter value C is incremented by 1 at R-6.

The output voltage vector V ^** and the counter value C are output by R-7.

By setting the counter value C of the command voltage vector matrix to the output voltage vector V ^* while looping the counter values in this way, one voltage vector can be read out from the voltage vectors for a plurality of cycles.

However, it is assumed that the timing when the input counter previous value Cz is 1 and the timing when the command voltage vector matrix V ^* is updated are synchronized. Note that if this cannot be observed, the output voltage will be different from what the prediction unit 7 assumes, which will lead to a prediction error.

As shown above, according to the third embodiment, the current change during the calculation time is predicted based on FIGS. 8, 9 and 10, and the voltage vector for a plurality of cycles is determined by one prediction calculation. By performing the MPC, it is possible to perform control that achieves high current control performance, low switching frequency, and securement of a calculation time longer than that of the first and second embodiments.

In addition, there is an advantage that the prior literature considers the case where the voltage vector for a plurality of cycles is determined by the current prediction calculation, which is applicable when the prediction period is general.

[Embodiment 4]
FIG. 11 shows a configuration diagram of MPC2 according to the fourth embodiment. In the fourth embodiment, a configuration in which the timing shift period of the output voltage change with respect to the current detection is taken into consideration and a configuration in which the prediction unit 7 outputs voltage vectors for a plurality of cycles are combined. That is, FIG. 6 of the second embodiment and FIG. 8 of the third embodiment are combined.

The third embodiment has a configuration corresponding to the case where the voltage vector for a plurality of cycles is handled in the first embodiment. The third embodiment can be used in combination with the second embodiment, which is referred to as the fourth embodiment.

As shown in FIG. 11, the output of the prediction unit 7 is set to a command voltage vector matrix V ^* which is a voltage vector for a plurality of output cycles, a read-out unit 9 is provided in the subsequent stage, and the prediction unit 7 has a value two times before the command voltage vector matrix. Vzz ^* is substituted. It is assumed that the reading unit 9 operates in the same manner as in FIG. 10 of the third embodiment.

FIG. 12 shows a flowchart of the prediction unit 7 in the fourth embodiment. FIG. 12 is a combination of the functions of the flowcharts of FIGS. 7 and 9, that is, the second and third embodiments.

The operation of the flowchart of the prediction unit 7 shown in FIG. 12 will be described. In FIG. 12, before calculating the current change amounts Δidc and Δiqc during the calculation time, the current change amounts Δiddel and Δiqdel during the delay time Tdel are considered.

In 4-1 the command voltage vector matrix two times before the previous value Vzz ^* is added and input to 3-1 in FIG.

In 4-2, the current change amounts Δiddel and Δiqdel in the delay time Tdel are obtained for the Ncalcth voltage vector which is the final output cycle of the command voltage vector matrix two times before the previous value Vzz ^*. Based on the equation (3), the calculation may be performed as in the equation (22). However, vdzz1 and vqzz1 are conversion values of the command voltage vector matrix two times before the previous value Vzz ^* [Ncalc] to the dq-axis voltage.

In 4-3, the current id'and iq', which are the sums of the detected d-axis current id and the detected q-axis current iq and the current changes Δiddel and Δiqdel, are set as the initial values of repetition in consideration of the current change during the calculation time. .. This means that the delay time Tdel of the output change from the detection is taken into consideration.

By the way, it is assumed that the delay time Tdel is equal to or less than the output cycle Tc, and only the Ncalcth from the ^{command voltage vector matrix pre-previous value Vzz * is used.} In this regard, in order to reduce the input information, only the Ncalcth value may be input instead of all the ^{command voltage vector matrix pre-previous value Vzz *.}

Further, when the delay time Tdel is equal to or longer than the output cycle Tc, the processing may be divided into a plurality of processes as in the case of repeating the current change amount calculation during the calculation time. For example, in the case of the condition of the equation (23), if the equations (24) to (29) are processed in the order of the equation numbers, the current change amounts Δiddel and Δiqdel in the delay time Tdel can be considered. However, vdzz2 and vqzz2 are conversion values of Vzz * [Ncalc-1] to the dq-axis voltage.

4-4 to 4-17 are the same as 3-3 to 3-16 in FIG.

As shown above, according to the fourth embodiment, the second and third embodiments are combined, that is, the amount of change in current during the calculation time is predicted and the amount of change during the calculation time is predicted based on FIGS. 10, 11, and 12. High current control performance, low switching frequency,

Embodiments

1 and 2 by determining the voltage vector for a plurality of cycles with one prediction calculation and performing MPC in consideration of the timing difference between current detection and output voltage change. It is possible to secure a longer calculation time and perform control that achieves current detection at a timing that is not affected by ringing.

Further, for the prior art, when the voltage vector for a plurality of cycles is determined by the current prediction calculation, which is applicable when the prediction period is general, the difference between the detection time and the output voltage change time is taken into consideration. There is an advantage that it considers.

Although the above description has been made in detail only with respect to the specific examples described in the present invention, it is clear to those skilled in the art that various modifications and modifications can be made within the scope of the technical idea of the present invention. It goes without saying that such modifications and modifications fall within the scope of the claims.

Claims

A host control unit that generates a command current based on the command value,
By model prediction control, a plurality of assumed voltage vectors are set for each output cycle during the prediction period, and the predicted current of the assumed voltage vector is calculated based on the correction current value obtained by adding the amount of current change during the calculation cycle to the detected current. Then, the evaluation function of the assumed voltage vector and the predicted current is calculated, and the combination of the assumed voltage vector having the highest evaluation function is determined from the combination of the assumed voltage vectors in the prediction period. Is a predictor that outputs the assumed voltage vector that is actually output from the combination of the assumed voltage vectors at the highest level as the command voltage vector, and
A gate signal determination unit that outputs a gate signal for outputting the voltage represented by the command voltage vector from the power converter, and a gate signal determination unit.
A power converter that is driven and controlled based on the gate signal,
Power conversion system with.
The power conversion system according to claim 1, wherein the amount of change in current during the calculation cycle is calculated by the following equation (5).

Δidc: Current change amount during calculation cycle (d-axis)
Δiqc: Current change amount during calculation cycle (q-axis)
Tc: Output cycle (= calculation cycle)
id: Detected d-axis current iq: Detected q-axis current vdz: d-axis voltage vector Previous value vqz: q-axis voltage vector Previous value R: Winding resistance Ld: d-axis inductance Lq: q-axis inductance ωr: Detected electrical angular velocity ψ: Permanent magnet interlinkage magnetic flux number
A host control unit that generates a command current based on the command value,
By model prediction control that determines the voltage vector of multiple output cycles in one calculation cycle, multiple assumed voltage vectors are set for each output cycle during the prediction period, and the correction current is added to the detected current by the amount of current change during the calculation cycle. The predicted current of the assumed voltage vector is calculated based on the value, the evaluation function of the assumed voltage vector and the predicted current is calculated, and the evaluation function is the highest among the combinations of the assumed voltage vectors in the prediction period. A predictor that outputs the combination of the assumed voltage vectors as a command voltage vector matrix, and
A reading unit that selects the voltage vector of the output cycle this time from the command voltage vector matrix of multiple output cycles and outputs it as an output voltage vector.
A gate signal determination unit that outputs a gate signal for outputting the voltage represented by the output voltage vector from the power converter, and a gate signal determination unit.
A power converter that is driven and controlled based on the gate signal,
Power conversion system with.
The current change amount during the calculation cycle is a value obtained by calculating the current change amount during each output cycle by the following equation (21) and adding the current change amounts of all the output cycles during the calculation cycle. The power conversion system described.

Δidc: Current change amount during output cycle (d-axis)
Δiqc: Current change amount during output cycle (q-axis)
Tc: Output cycle id': Detected d-axis current or d-axis current iq': Detected q-axis current or q-axis current vdz: d-axis voltage vector Previous value vqz: q-axis voltage vector Previous value R: Winding Resistance Ld: d-axis inductance Lq: q-axis inductance ωr: Detected electrical angular velocity ψ: Permanent magnet interlinkage magnetic flux number
The power conversion system according to claim 1 or 3, wherein the correction current value is a value obtained by adding the amount of change in the current during the calculation cycle and the amount of change in the current in the timing shift period of the output voltage change with respect to the current detection to the detected current. ..
The power conversion system according to claim 5, wherein the amount of current change in the timing shift period of the output voltage change with respect to the current detection is calculated by the following equation (17).

Δiddel: Current change amount (d-axis) during the timing shift period of the output voltage change with respect to the current detection
Δiqdel: Current change amount (q-axis) during the timing shift period of the output voltage change with respect to the current detection
Tdel: Timing shift period of output voltage change with respect to current detection id: Detected d-axis current iq: Detected q-axis current vdzz: d-axis voltage vector Pre-previous value vqz: q-axis voltage vector Pre-previous value R: Winding resistance Ld: d-axis inductance Lq: q-axis inductance ωr: detected electric angular velocity ψ: permanent magnet interlinkage magnetic flux number
A control method for a power conversion system equipped with a power converter that is driven and controlled based on a gate signal.
In the upper control unit, a command current is generated based on the command value,
In the prediction unit, a plurality of assumed voltage vectors are set for each output cycle during the prediction period by model prediction control, and the assumed voltage vector is calculated based on the correction current value obtained by adding the amount of change in the current during the calculation cycle to the detected current. The predicted current is calculated, the assumed voltage vector and the evaluation function of the predicted current are calculated, and the combination of the assumed voltage vector having the highest evaluation function is determined from the combinations of the assumed voltage vectors in the predicted period. , The assumed voltage vector actually output from the combination of the assumed voltage vectors at which the evaluation function is the highest is output as the command voltage vector.
A control method of a power conversion system that outputs a gate signal for outputting a voltage represented by the command voltage vector from the power converter in the gate signal determination unit.
A control method for a power conversion system equipped with a power converter that is driven and controlled based on a gate signal.
In the upper control unit, a command current is generated based on the command value,
In the prediction unit, a plurality of assumed voltage vectors are set for each output cycle during the prediction period by model prediction control that determines voltage vectors for multiple output cycles in one calculation cycle, and the amount of current change during the calculation cycle is set as the detection current. The predicted current of the assumed voltage vector is calculated based on the added correction current value, the evaluation function of the assumed voltage vector and the predicted current is calculated, and the evaluation function is selected from the combination of the assumed voltage vectors in the prediction period. Outputs the combination of the assumed voltage vectors at the highest order as a command voltage vector matrix.
In the reading unit, the voltage vector of the output cycle this time is selected from the command voltage vector matrix of a plurality of output cycles and output as an output voltage vector.
A control method of a power conversion system that outputs a gate signal for outputting a voltage represented by the output voltage vector from the power converter in the gate signal determination unit.