Active control type magnetic levitation-based system
The magnetic levitation system uses an MPC to evaluate current and voltage differences between H-bridge circuits, improving responsiveness and optimizing switching frequency for precise control.
Patent Information
- Application Number
- JP2023082777
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-19
- Publication Date
- 2024-11-29
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing magnetic levitation systems using H-bridge circuits lack a clear model predictive control (MPC) configuration suitable for evaluating current differences between opposing electromagnets, which is crucial for controlling the attractive force in magnetic levitation systems.
The system employs an MPC that evaluates current and voltage differences between first and second H-bridge circuits, using state machines to determine optimal gate signals based on output voltage states, thereby controlling the magnetic levitation system effectively.
This approach enhances the responsiveness and optimizes switching frequency, ensuring precise control of the magnetic levitation system by predicting and evaluating current changes, reducing pulsations in the attractive force.
Smart Images

Figure 2024166588000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to magnetic levitation and magnetic bearings. [Background technology]
[0002] In an active control magnetic levitation system, a current is passed through the electromagnet windings to generate a magnetic attraction force on the levitated body, thereby controlling the levitated body to a desired displacement.
[0003] H-bridge circuits are widely used as converters for magnetic levitation systems, especially for large capacity applications. Unlike general step-down choppers, H-bridge circuits can apply negative voltage between the winding terminals, and therefore have superior current response.
[0004] The current of the electromagnet winding is controlled by a converter, and the most common control method is a combination of PI control (proportional-integral control) and triangular wave comparison PWM (pulse width modulation). In contrast to this, there is also a method called MPC (model predictive control).
[0005] MPC predicts the current change caused by the output voltage at future times for many switching patterns, and actually outputs the switching pattern that results in the optimal current trajectory. MPC has the advantages of being able to follow step fluctuations well and to optimize the switching frequency according to the operating state. [Prior art documents] [Patent documents]
[0006] [Patent Document 1] JP 2018-74751 A [Patent Document 2] JP 2018-74745 A [Patent Document 3] JP 2021-170882 A Summary of the Invention [Problem to be solved by the invention]
[0007] Regarding MPC in converters, the following prior art is disclosed.
[0008] Patent Documents 1 and 2 disclose an MPC that determines the duty ratio for the next cycle in a synchronous rectification chopper that generates a DC voltage.
[0009] Patent Document 3 discloses an MPC that determines output voltage vectors for the next several cycles in a three-phase inverter for driving a motor.
[0010] It can be seen that the above prior art is directed to a three-phase inverter and a synchronous rectification chopper.
[0011] In MPC, the assumed model is important for predicting the future state of the model, but the model form of MPC suitable for H-bridge circuits and its realization configuration have not been clarified.For example, in an H-bridge circuit, it is more advantageous in terms of calculation to state the output voltage and determine the gate signal in a separate process than to directly determine the gate signal, but the configuration has not been clarified.
[0012] In addition, in the MPC of motor drive inverters up to now, the current and rotation speed have mainly been introduced into the evaluation function. In the MPC of synchronous rectification choppers, the current and capacitor voltage have mainly been introduced into the evaluation function. On the other hand, in magnetic levitation systems, the difference in the attractive forces of two opposing electromagnets is important for braking the levitated body. In other words, it is important to evaluate the current difference between the two converters, but this point has not been considered in prior art.
[0013] For the reasons mentioned above, the challenge is to provide an actively controlled magnetic levitation system that uses MPC. [Means for solving the problem]
[0014] The present invention has been devised in view of the above-mentioned problems in the related art, and one aspect of the present invention is characterized by comprising first and second electromagnets provided at positions opposing a levitation body, first and second circuits for controlling a first electromagnet winding current and a second electromagnet winding current, a DC power supply which serves as a power supply for the first and second circuits, and an MPC which evaluates a first circuit current evaluation value, a first circuit voltage evaluation value, a second circuit current evaluation value, a second circuit voltage evaluation value, and a current difference evaluation value between the first circuit and the second circuit for an output voltage from the next cycle onwards, taking into account models of the first circuit, the winding of the first electromagnet, and the second circuit, and outputs a gate signal of the first circuit and a gate signal of the second circuit which correspond to the output voltage with the best evaluation result.
[0015] In one aspect of the invention, the first circuit is a first H-bridge circuit, the second circuit is a second H-bridge circuit, and when the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit output a positive DC power supply voltage, the state is 1, when the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit output a positive DC power supply voltage, the state is 2, when the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit output a negative DC power supply voltage, the state is 3, the pattern of the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit is one in the case of state 1, two in the case of state 2, and one in the case of state 3, transition from state 1 to state 3 and from state 3 to state 1 are prohibited, when transitioning from state 1 or state 3 to state 2 and when transitioning from state 2 to state 1 or state 3, the pattern of the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit are switched, the pattern of the gate signal is not switched while in state 2, and the pattern of the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit used in state 2 is a pattern different from the pattern of the previous state 2.
[0016] In another aspect, the first circuit is a first H-bridge circuit, and the second circuit is a second H-bridge circuit; when the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit output a positive DC power supply voltage, the state is 3; when the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit output a positive DC power supply voltage, the state is 2; when the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit output a negative DC power supply voltage, the state is 1; the pattern of the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit is one in the case of state 1, two in the case of state 2, and one in the case of state 3; transition from state 1 to state 3 and from state 3 to state 1 are prohibited; when transitioning from state 1 or state 3 to state 2 and when transitioning from state 2 to state 1 or state 3, the patterns of the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit are switched, and the pattern of the gate signal is not switched while in state 2; and the pattern of the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit used in state 2 is a pattern different from the pattern of the previous state 2.
[0017] In one aspect, the MPC calculates the first circuit current evaluation value and the second circuit current evaluation value using equations (21) and (22), calculates the first circuit voltage evaluation value and the second circuit voltage evaluation value using equation (23), and calculates the current difference evaluation value using equations (24) and (25).
[0018]
number
[0019]
number
[0020]
number
[0021]
number
[0022]
number
[0023] Ji_j: The first circuit current evaluation value and the second circuit current evaluation value at any time j G_Ji: Current evaluation gain i_j: Output current at any time j i_ref: Command current G_Jiabs: Deviation penalty gain i_abs_ref: Current deviation tolerance Jv_j: First circuit voltage evaluation value, second circuit voltage evaluation value at any time j v_j: The state of the output voltage at any time j v_j+1: State of the output voltage at time j+1 G_Jv: Voltage rating gain G_JNG: Forbidden transition penalty gain JiF_j: Current difference evaluation value G_JiF: Current difference evaluation gain iA-B_j(=iF_j): Output current difference at any time j iA-B_ref(=iF_ref): Command current difference G_JiFabs: Differential deviation penalty gain iF_abs_ref: Current difference deviation tolerance.
[0024] In another aspect, the MPC calculates the first circuit current evaluation value and the second circuit current evaluation value using equations (28) and (29), calculates the first circuit voltage evaluation value and the second circuit voltage evaluation value using equation (23), and calculates the current difference evaluation value using equations (30) and (31).
[0025]
number
[0026]
number
[0027]
number
[0028]
number
[0029]
number
[0030] Ji_j: The first circuit current evaluation value and the second circuit current evaluation value at any time j G_Ji: Current evaluation gain i_j: Output current at any time j i_ref: Command current G_Jiabs: Deviation penalty gain i_abs_ref: Current deviation tolerance Jv_j: First circuit voltage evaluation value, second circuit voltage evaluation value at any time j v_j: The state of the output voltage at any time j v_j+1: State of the output voltage at time j+1 G_Jv: Voltage rating gain G_JNG: Forbidden transition penalty gain JiF_j: Current difference evaluation value G_JiF: Current difference evaluation gain iA-B_j(=iF_j): Output current difference at any time j iA-B_ref(=iF_ref): Command current difference G_JiFabs: Differential deviation penalty gain iF_abs_ref: Current difference deviation tolerance.
[0031] In one aspect, the MPC includes a current change estimation unit that estimates a first electromagnet winding current at a currently determined output start time and a second electromagnet winding current at the currently determined output start time; an MPC evaluation calculation unit that outputs data on output voltages of the first H-bridge circuit and the second H-bridge circuit which have the best evaluation results from the currently determined output start time to a prediction calculation consideration period, based on the first electromagnet winding current at the currently determined output start time, the second electromagnet winding current at the currently determined output start time, the first circuit command current of Equation (7), and the second circuit command current of Equation (8); and a gate determination unit that outputs the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit, based on the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit.
[0032]
number
[0033]
number
[0034] iA_ref: 1st circuit command current iB_ref: 2nd circuit command current ibias: command bias current iFref: Command attraction force control current.
[0035] As one aspect of the present invention, the MPC evaluation calculation unit predicts and evaluates a current change for each pattern of the output voltage in one cycle, sorts the evaluation results in order of best evaluation result each time prediction and evaluation is performed for one cycle, selects a selected number of data with best evaluation results, predicts and evaluates a current change for each pattern of the output voltage in the next one cycle for each of the selected number of evaluation results, repeats the prediction and evaluation of the current change for each pattern of the output voltage for one cycle and the sorting and selection of the evaluation results, and performs prediction and evaluation up to the prediction calculation consideration cycle, and outputs data regarding the output voltage of the first H-bridge circuit and the second H-bridge circuit which has the best evaluation result for the prediction calculation consideration cycle.
[0036] In one embodiment, the MPC evaluation calculation unit sorts and selects the evaluation results by sorting and selecting the input signals at once using a sorting network.
[0037] In another aspect, the MPC evaluation calculation unit is characterized in that, when sorting and selecting evaluation results, the sorting network is multi-staged, input signals are divided into multiple groups, sorted and selected by multiple sorting networks, and the selected signals are sorted and selected by a sorting network of the next stage.
[0038] In one aspect, the current change estimation unit uses a current estimation calculation unit that estimates the current one cycle ahead to calculate an estimate of the output current at time j+1 using equations (14) and (15), and repeats this calculation until time j+1 becomes the currently determined output start time, thereby obtaining the output current at the currently determined output start time.
[0039]
number
[0040]
number
[0041] i_j+1: Output current at time j+1 i_j: Output current at any time j Tc: Output voltage switching period L_load: Winding inductance Tdead: Dead time VH_j: The output voltage of the H-bridge circuit corresponding to the state of the output voltage at any time j VH_d: Output voltage during dead time R_load: Winding resistance VH_in: Positive DC power supply voltage -VH_in: Negative DC power supply voltage v_j: The state of the output voltage at any time j.
[0042] In one aspect, the output voltage of the first H-bridge circuit during the dead time is determined as follows: if the polarity of the first electromagnet winding current is positive and the state value is decreased by one from the previous value, the state during the dead time is a state before the transition; if the polarity of the first electromagnet winding current is positive and the state value is increased by one from the previous value, the state during the dead time is a state after the transition; if the polarity of the first electromagnet winding current is negative and the state value is increased by one from the previous value, the state during the dead time is a state before the transition; and if the polarity of the first electromagnet winding current is negative and the state value is decreased by one from the previous value, the state during the dead time is a state after the transition. and determining an output voltage of the second H-bridge circuit during the dead time such that, when the polarity of the second electromagnet winding current is positive and the state value is decreased by one from the previous value, the state is set to a state before the transition during the dead time, when the polarity of the second electromagnet winding current is positive and the state value is increased by one from the previous value, the state is set to a state after the transition during the dead time, when the polarity of the second electromagnet winding current is negative and the state value is increased by one from the previous value, the state is set to a state before the transition during the dead time, and when the polarity of the second electromagnet winding current is negative and the state value is decreased by one from the previous value, the state is set to a state after the transition during the dead time. Effect of the Invention
[0043] According to the present invention, it is possible to provide an active control type magnetic levitation system in which control is performed by MPC. [Brief description of the drawings]
[0044] [Figure 1] FIG. 1 is a block diagram showing an active control type magnetic levitation system. [Diagram 2] FIG. 4 is an explanatory diagram showing forces acting on a levitating body. [Diagram 3] MPC control configuration diagram of an active control magnetic levitation system. [Figure 4]A diagram showing the gate signal pattern and output voltage of an H-bridge circuit. [Diagram 5] FIG. 1 shows the state definition of the output voltage. [Figure 6] FIG. 2 is a configuration diagram showing an MPC according to the first embodiment. [Figure 7] A diagram showing the relationship between MPC calculations and sampling / output. [Figure 8] FIG. 4 is a diagram showing the definition of time in the MPC of the first embodiment. [Figure 9] FIG. 4 is a configuration diagram showing a gate determination unit. [Figure 10] FIG. 4 is a diagram showing a current change estimation unit. [Figure 11] FIG. [Figure 12] FIG. 4 is a diagram showing a dead-time voltage estimating unit. [Figure 13] FIG. 2 is a diagram showing the current path and output voltage during a dead time period. [Figure 14] FIG. 2 is a diagram showing an MPC evaluation calculation unit. [Figure 15] FIG. 13 is a diagram showing a one-cycle prediction evaluation unit. [Figure 16] FIG. [Figure 17] FIG. 2 is a diagram showing an evaluation calculation unit according to the first embodiment. [Figure 18] FIG. 2 is a diagram showing an Nin input, Ns output sort selection unit of the first embodiment. [Figure 19] A diagram showing an example of the operation of a five-input sorting network. [Figure 20] FIG. [Figure 21] FIG. 11 is a diagram showing a sort selection unit according to the second embodiment. [Figure 22] FIG. 11 is a diagram showing an evaluation calculation unit according to the third embodiment. [Diagram 23] FIG. 4 is a diagram showing a simulation result of the first embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0045] Hereinafter, first to third embodiments of an active control type magnetic levitation system according to the present invention will be described in detail with reference to FIGS.
[0046] [Embodiment 1] Figure 1 shows the configuration of an active control magnetic levitation system. In this system, a first electromagnet 2A and a second electromagnet 2B are provided in positions facing a levitation body 1, and their winding currents are controlled by a first circuit (e.g., an H-bridge circuit, hereafter referred to as the first H-bridge circuit) 3A and a second circuit (e.g., an H-bridge circuit, hereafter referred to as the second H-bridge circuit) 3B, respectively. A DC power supply 4 is provided as the power source for the first H-bridge circuit 3A and the second H-bridge circuit 3B. The DC power supply 4 may be a battery or the output voltage of an AC / DC converter connected to an AC power supply system, or any other power source generated in a desired configuration.
[0047] The first H-bridge circuit 3A is composed of a capacitor CA and switching elements (e.g., MOSFETs: Metal Oxide Semiconductor Field Effect Transistors) SA1, SA2, SA3, and SA4. Similarly, the second H-bridge circuit 3B is composed of a capacitor CB and switching elements SB1, SB2, SB3, and SB4.
[0048] Additionally, VH_in is the DC power supply voltage [V], iA_out is the first electromagnet winding current [A], iB_out is the second electromagnet winding current [A], VH_A is the first H-bridge circuit output voltage [V], and VH_B is the second H-bridge circuit output voltage [V], each defined in the direction shown in Figure 1.
[0049] In this system, a first electromagnet 2A and a second electromagnet 2B are disposed facing the levitation body 1, and the displacement of the levitation body 1 is controlled by the attractive forces of these. However, the configuration is not limited to that shown in Fig. 1. For example, the switching elements SA1 to SA4 and SB1 to SB4 of the first and second H-bridge circuits 3A and 3B do not have to be MOSFETs, and for example, IGBTs (insulated gate bipolar transistors) may be used.
[0050] Figure 2 shows the forces acting on the levitation body 1. If the attractive force of the first electromagnet 2A is FA (the direction of the first electromagnet 2A is positive), the attractive force of the second electromagnet 2B is FB (the direction of the second electromagnet 2B is positive), the steady disturbance force on the levitation body 1 due to gravity and other factors is Fd (the direction of the second electromagnet 2B is positive), and the resultant force acting on the levitation body 1 is Flev (the direction of the first electromagnet 2A is positive), the relationship between them is expressed by equation (1).
[0051]
number
[0052] Here, under linearization around the equilibrium point, the attractive forces of the first electromagnet 2A and the second electromagnet 2B are expressed by equations (2) and (3).
[0053]
number
[0054]
number
[0055] ΔiA is the amount of current change [A] from the equilibrium point of the first electromagnet 2A, ΔiB is the amount of current change [A] from the equilibrium point of the second electromagnet 2B, Ki is the current-attraction force gain [F / A], Kd is the displacement-attraction force gain [F / m], and Δd is the displacement [m] from the equilibrium point between the levitating body 1 and the first electromagnet 2A, and the second electromagnet 2B.
[0056] Considering that the electromagnetic forces of the first and second electromagnets 2A, 2B are in the attraction direction regardless of the current polarity, in this embodiment 1, iA_out≧0, iB_out≧0. The direction of displacement Δd approaching the first electromagnet 2A is set to be positive. Also, it is assumed that the current-attraction force gain Ki and the displacement-attraction force gain Kd of the first and second electromagnets 2A, 2B are the same, and any deviation appears as a disturbance.
[0057] Substituting equations (2) and (3) into equation (1), we obtain equation (4).
[0058]
number
[0059] This means that it is possible to control the displacement of the magnetic levitation system by appropriately controlling the current difference (ΔiA-ΔiB) between the first and second electromagnets 2A and 2B, which are installed opposite each other.
[0060] In actual operation, in order to use the electromagnets in the linear region, a bias current is passed through both the first electromagnet 2A and the second electromagnet 2B, and a current for controlling the attractive force is added to the first electromagnet 2A and the second electromagnet 2B with opposite polarity and 0.5 times the current.
[0061] The displacement control of the levitation body 1 is assumed to be performed separately in a desired manner, and the present embodiment 1 deals with the current control after the command currents for the two electromagnetic windings disposed opposite to each other have been determined.
[0062] In the first embodiment, MPC is used for current control of the first and second H-bridge circuits 3A and 3B in order to improve responsiveness and optimize the switching frequency of the switching elements.
[0063] MPC predicts in advance how the future state of the controlled object will change in relation to future output for many patterns, and actually outputs the output that gives the best prediction result. This allows optimal control to be performed within the range of the MPC's output resolution and sampling resolution, provided that the model and plant match.
[0064] The MPC control configuration of the magnetic levitation system is shown in Figure 3. It shows a feedback control configuration in which the winding currents of the first electromagnet 2A and the second electromagnet 2B are controlled by the first H-bridge circuit 3A and the second H-bridge circuit 3B. The DC power supply 4, the first H-bridge circuit 3A, the second H-bridge circuit 3B, the first electromagnet 2A, and the second electromagnet 2B are configured in the same way as in Figure 1.
[0065] A command bias current ibias [A] and a command attractive force control current iFref [A] are input to the MPC 5. These commands are determined by a higher-level displacement control so as to maintain the displacement of the levitation body 1 as appropriate. In addition, a first electromagnet winding current iA_out [A] and a second electromagnet winding current iB_out [A] are input to the MPC 5. The first electromagnet winding current iA_out [A] and the second electromagnet winding current iB_out [A] are detected by a desired method such as a Hall current sensor or a shunt resistor.
[0066] The MPC 5 evaluates the output voltage from the next cycle onwards, taking into account the models of the H-bridge circuit and the electromagnetic winding. For the output voltage with the best evaluation result, a gate signal corresponding to that output voltage is output as a gate signal gA for the first H-bridge circuit 3A and a gate signal gB for the second H-bridge circuit 3B. In the first H-bridge circuit 3A and the second H-bridge circuit 3B, switching elements are driven based on these gate signals gA and gB.
[0067] 3 shows the overall configuration of current control by the MPC 5. The first electromagnet winding current iA_out and the second electromagnet winding current iB_out are fed back to determine gate signals gA and gB for controlling them to MPC command currents, and the first H-bridge circuit 3A and the second H-bridge circuit 3B are driven.
[0068] Here, the first H-bridge circuit 3A includes four switching elements SA1 to SA4. The gate signals of these four switching elements SA1, SA2, SA3, and SA4 are denoted as gA1, gA2, gA3, and gA4, respectively, and the gate signal gA is defined by equation (5).
[0069]
number
[0070] Similarly, in the second H-bridge circuit 3B, the gate signals of the four switching elements SB1, SB2, SB3, and SB4 are gB1, gB2, gB3, and gB4, respectively, and the gate signal gB is defined by equation (6).
[0071]
number
[0072] For the gate signals, the ON / OFF state is determined by 0 and 1. In other words, when (gA1, gA2, gA3, gA4) is (1, 0, 0, 1), gA1 is ON, gA2 is OFF, gA3 is OFF, and gA4 is ON.
[0073] Based on this, we consider an MPC configuration suitable for an H-bridge circuit.
[0074] Figure 4 shows the gate signal pattern and output voltage of the H-bridge circuit. It shows the current path and the value of the first H-bridge circuit output voltage VH_A according to the combination of (gA1, gA2, gA3, gA4) in the first H-bridge circuit 3A. Considering that turning on one leg (the combination of SA1 and SA2, or SA3 and SA4) at the same time shorts out the DC voltage, and turning off one leg at the same time creates a diode path and increases losses, the gate signal combinations that are actually used are limited to the four types shown in Figure 4.
[0075] Here, there are four types of gate signal combinations (patterns), while there are three types of output voltage: positive DC power supply voltage VH_in, 0, and negative DC power supply voltage -VH_in, and (1,0,1,0) and (0,1,0,1) are the same zero voltage output. Considering the equalization of the load on the switching element, it is not permissible to use only one of them, and the zero voltage output requires that the gate signals (1,0,1,0) and (0,1,0,1) be used as evenly as possible.
[0076] In addition, (1,0,0,1) and (0,1,1,0), and (1,0,1,0) and (0,1,0,1) are inverted gate signals, and to transition between these, it is necessary to turn all switching elements ON / OFF at once. This is undesirable from the viewpoint of reducing the safety of gate drive or complicating gate drive control to ensure safety.
[0077] In consideration of the above, in the first embodiment, the MPC is configured based on the concept of a state machine (state transition) based on the first H-bridge circuit output voltage VH_A.
[0078] Figure 5 shows the state definitions of the output voltage. Although the first H-bridge circuit 3A is shown, the second H-bridge circuit 3B is similar and the only difference is replacing the symbol A with B, so the diagram is omitted. Also, this state definition is not a state machine in the strict sense, and there are no specific transition conditions. This is because in MPC, the output voltage is determined by evaluating the prediction results, and transition conditions cannot be determined. Instead, the relationship between each output voltage is described with transition allowed and transition prohibited.
[0079] State 1 (vA=1) is a state in which a positive DC power supply voltage VH_in is output between the load terminals. The gate signal at this time is (1, 0, 0, 1).
[0080] State 2 (vA=2) is a state in which zero voltage is output between the load terminals. The gate signals at this time are (1,0,1,0), (0,1,0,1). The gate signals do not change while in state 2, but when transitioning from state 2 to state 1 or state 3, the gate signal used in state 2 is switched to one different from the previous one.
[0081] This prohibits the transitions (1,0,1,0) and (0,1,0,1) among the relationships in which all gate signals are inverted, and uses (1,0,1,0) and (0,1,0,1) as evenly as possible. The detailed configuration of the gate signal processing for State 2 will be shown later in the gate determination section.
[0082] State 3 (vA=3) is a state in which a negative DC power supply voltage −VH_in is output between the load terminals. The gate signal at this time is (0, 1, 1, 0).
[0083] States 1 and 2 can transition to each other, and states 2 and 3 can transition to each other, but transitions between states 1 and 3 are prohibited. This prohibits transitions between (1,0,0,1) and (0,1,1,0) when all gate signals are inverted. In this state machine, it can be simply defined that if the difference between the first circuit output voltage and the previous value of state vA is 1 or less, then the transition is permitted, and if it is 2, the transition is prohibited.
[0084] If the output voltage of MPC5 is directly referenced to the gate signal without using the state vA of the first circuit output voltage, it is necessary to consider which of the two types of gate signals corresponds to the zero output voltage at the prediction and evaluation stage of MPC5, which complicates the case distinction. Also, MPC has many identical prediction and evaluation configurations. If the case distinction of the gate signal in each prediction and evaluation becomes complicated, the same configuration will be used many times overall, which will cause a burden in implementation.
[0085] However, in this state machine, the state vA of the first circuit output voltage is made variable. The value of the state vA of the first circuit output voltage and the output voltage value correspond uniquely, and the gate determination unit in the subsequent stage determines which of the two types of gate signals corresponds to the zero output voltage, which simplifies the prediction and evaluation configuration.
[0086] As described above, by defining the state based on the output voltage in the H-bridge circuit, it is possible to achieve a configuration that is more advantageous in terms of implementation than a configuration that directly references the gate signal.
[0087] However, in consideration of enabling equivalent operation, the definitions of states 1 and 3 in this embodiment 1 may be interchanged. Regarding state 2, in order to prohibit transitions between (1,0,0,1) and (0,1,1,0), if the difference between the first circuit output voltage state vA and the previous value is 2, transition is prohibited, and therefore it cannot be interchanged with other states.
[0088] The state vB of the second circuit output voltage is also considered to be an output voltage reference state, similar to the state vA of the first circuit output voltage. Even if a symbol referring to time is added to the state vA_j of the first circuit output voltage at time j, it is treated in the same way as this state definition.
[0089] FIG. 6 shows a configuration diagram of the MPC 5 according to the first embodiment.
[0090] A multiplier 6 multiplies the command attraction force control current iFref by 1 / 2. An adder 7 adds the output of the multiplier 6 to the command bias current ibias and outputs the result as a first circuit command current iA_ref. A subtractor 8 subtracts the output of the multiplier 6 from the command bias current ibias and outputs the result as a second circuit command current iB_ref.
[0091] The current change estimation unit 9A receives the first electromagnet winding current iA_out and the first circuit previous voltage array VA_z. The current change estimation unit 9B receives the second electromagnet winding current iB_out and the second circuit previous voltage array VB_z. The current change estimation units 9A and 9B output the first electromagnet winding current iA_k at time k and the second electromagnet winding current iB_k at time k.
[0092] The MPC evaluation calculation unit 10 receives the first circuit command current iA_ref, the second circuit command current iB_ref, the first electromagnet winding current iA_k at time k, the second electromagnet winding current iB_k at time k, the first circuit previous voltage array VA_z, and the second circuit previous voltage array VB_z. The MPC evaluation calculation unit 10 outputs the time k+Nc first-place data array Dk+Nc_1 which is data related to the output voltages of the first H-bridge circuit and the second H-bridge circuit having the best evaluation results from time k to the prediction calculation consideration period Nc.
[0093] The switching pattern information extraction unit 11 receives an input of the first-place data array Dk+Nc_1 at time k+Nc. The switching pattern information extraction unit 11 outputs a first circuit evaluation value minimum voltage array VA_Jmin and a second circuit evaluation value minimum voltage array VB_Jmin from the information contained in the first-place data array Dk+Nc_1 at time k+Nc. The first circuit evaluation value minimum voltage array VA_Jmin and the second circuit evaluation value minimum voltage array VB_Jmin are input to the current change estimation units 9A, 9B and the MPC evaluation calculation unit 10 as the first circuit previous voltage array VA_z and the second circuit previous voltage array VB_z delayed by one control period in buffers 12A, 12B.
[0094] The current output selection unit 13 receives the first circuit evaluation value minimum voltage array VA_Jmin and the second circuit evaluation value minimum voltage array VB_Jmin as input, and outputs the first circuit current output voltage vA_out and the second circuit current output voltage vB_out.
[0095] The gate determination units 14A and 14B receive the first circuit current output voltage vA_out and the second circuit current output voltage vB_out as inputs, and output the gate signal gA and the gate signal gB, respectively.
[0096] First, the first circuit command current iA_ref and the second circuit command current iB_ref are calculated based on the command bias current ibias and the command attraction force control current iFref. An appropriate calculation formula for this depends on the definition of the command attraction force control current iFref, but in the first embodiment, the command attraction force control current iFref is defined as the current difference (iA_out-iB_out) between the first electromagnet winding current and the second electromagnet winding current. Therefore, the first circuit command current iA_ref and the second circuit command current iB_ref are given by equations (7) and (8).
[0097]
number
[0098]
number
[0099] The first circuit command current iA_ref and the second circuit command current iB_ref obtained by this calculation are input to the MPC evaluation calculation unit 10.
[0100] Next, the estimation of the current change will be described.
[0101] Figure 7 shows the relationship between MPC calculations and sampling / output. It shows the time series relationship in which the current is sampled, MPC calculations are performed based on the current, and the output is determined. For simplicity, the diagram shows a configuration in which the MPC determines the output voltage for three cycles.
[0102] As shown in Figure 7, since the output voltage for multiple cycles is determined by MPC calculation, the previously determined output voltage is output during the period from the sampling time to when the output voltage determined by MPC calculation is actually output. The value sampled due to this previously determined output voltage changes during calculation. If the current used in the prediction calculation is deviated, the prediction result will also deviate, leading to a decrease in control performance. To prevent this, it is necessary to predict the current change for the period during which the previous output voltage is output, in other words, the period from the sampling time to when the output voltage determined this time actually starts to be output.
[0103] 8 shows the definition of time in the MPC 5 of the present embodiment 1. For simplification, the MPC 5 is illustrated with a configuration in which the output voltage is determined for three periods.
[0104] After a period from the sampling time during which the output voltage determined by the previous MPC has been output, the output voltage determined by the current MPC5 is output. At this time, the main role of the current MPC5 is to predict and evaluate the current change due to the currently determined output. For this reason, in this embodiment 1, the output start timing of the currently determined output is defined as time k. Under this definition, the sampling time is time k-3, and the time when the third output of the currently determined output is completed is time k+3.
[0105] In the configuration of the MPC 5 of the first embodiment, the number of periods is generalized, and the number of periods of the output voltage determined by the MPC 5 is (output voltage period number) No [period]. Then, the sampling time is k-No, and the time when the No-th output of the currently determined output is completed is time k+No.
[0106] As another definition of the period, the number of periods considered in the prediction calculation is set as (prediction calculation consideration period) Nc [periods]. The prediction calculation consideration period Nc does not necessarily have to match the number of output voltage periods No. For example, it is possible to predict the current up to 10 periods ahead by setting Nc=10 and No=2, and then actually output only the first two periods. However, if the prediction calculation consideration period Nc is smaller than the number of output voltage periods No, the output voltage for the number of output voltage periods No cannot be determined from the prediction results, so equation (9) becomes a constraint.
[0107]
number
[0108] Looking at FIG. 6 based on the above definitions of time, the first electromagnet winding current iA_out and the second electromagnet winding current iB_out are not input directly to the MPC evaluation calculation unit 10 but are input to current change estimation units 9A and 9B.
[0109] Here, as a measure against changes in sampled values during calculation, the current change up to time k, which is the start of output of the currently determined output, is estimated. At this time, information on the previous voltage is required, so the first circuit previous voltage array VA_z and the second circuit previous voltage array VB_z are also input to current change estimation units 9A, 9B. In addition, the outputs of current change estimation units 9A, 9B are the first electromagnet winding current iA_k at time k and the second electromagnet winding current iB_k at time k.
[0110] Voltage arrays such as the first circuit previous voltage array VA_z are arrays that have voltage state information for the prediction calculation consideration period Nc. This is defined by equation (10). For generalization, z, which indicates the previous value, has been omitted, but even if z is included, it is the same with each element simply adding z, such as vA_z1, vA_z2, etc. The same is true for the voltage of the second H-bridge circuit 3B, with only A and B being changed.
[0111]
number
[0112] The first circuit command current iA_ref, the second circuit command current iB_ref, and the first electromagnet winding current iA_k at time k and the second electromagnet winding current iB_k at time k, which are currents at the prediction / evaluation start timing, are input to the MPC evaluation calculation unit 10.
[0113] Using these, the MPC evaluation calculation unit 10 predicts the current change for each output voltage pattern in the previous cycle and performs evaluation using an evaluation function. Data related to the output voltages of the first and second circuits with the best evaluation results from time k to the prediction calculation consideration cycle Nc is output as the first data array Dk+Nc_1 at time k+Nc. The first circuit previous voltage array VA_z and the second circuit previous voltage array VB_z are also input to the MPC evaluation calculation unit 10, and are used when estimating the voltage during dead time.
[0114] The configuration of the gate determination unit 14 is shown in Fig. 9. This shows the inside of the gate determination units 14A and 14B in Fig. 6. This configuration is provided in multiple places with only the difference being the change in the input / output signals, so for generalization, the input / output names are given without distinguishing between the first circuit and the second circuit. The input / output signal names used in practice will be the same as those in Fig. 6. For example, v_out is replaced with vA_out and g is replaced with gA.
[0115] Comparator NEQ1 receives as input the current output voltage (state) v_out and constant 2, and outputs 1 if the current output voltage (state) v_out is not equal to constant 2, and outputs 0 otherwise. Comparator NEQ2 receives as input the current output voltage (state) v_out and a signal obtained by delaying the current output voltage (state) v_out by one control period through buffer 15, and outputs 1 if the current output voltage (state) v_out is not equal to the output of buffer 15, and outputs 0 otherwise.
[0116] The AND logic AND1 receives the outputs of the comparators NEQ1 and NEQ2, and outputs 1 if both are 1, and outputs 0 otherwise.
[0117] The selector SW1 receives as input a signal obtained by delaying the output VH0_sel of the selector SW1 by one control period in the buffer 16, and a signal obtained by passing the output signal of the buffer 16 through NOT logic NOT1. The selector SW1 also receives as input an output signal of AND logic AND1 as a selection signal in1, and outputs the output of the buffer 16 when the selection signal in1=0, and outputs the output signal of the NOT logic NOT1 when the selection signal in1=1. The output of the selector SW1 is VH0_sel.
[0118] The selector SW2 inputs the gate signal combinations (1,0,1,0) and (0,1,0,1) of the H-bridge circuit. The selector SW2 also inputs the output VH0_sel of the selector SW1 as the selection signal in2, and outputs (1,0,1,0) when the selection signal in2=0, and outputs (0,1,0,1) when the selection signal in2=1.
[0119] Selector SW3 inputs the output of selector SW2 and the gate signal combinations (1,0,0,1) and (0,1,1,0) of the H-bridge circuit. Selector SW3 also inputs the current output voltage (state) v_out as selection signal in3, and outputs (1,0,0,1) when selection signal in3=1, outputs the output signal of selector SW2 when selection signal in3=2, and outputs (0,1,1,0) when selection signal in3=3. The output of selector SW3 is the gate signal g' before dead time insertion.
[0120] The dead time 17 receives the gate signal g' before the dead time is inserted and outputs the gate signal g. The above is the configuration of the gate determination unit .
[0121] Figure 10 shows the current change estimation unit 9. This shows the inside of the current change estimation units 9A and 9B in Figure 6. As in Figure 9, for generalization, the input and output names are used without distinction between the first circuit and the second circuit.
[0122] The previous voltage array V_z is divided into output voltages (v_z1, v_z2, ..., v_zNo) for each period it contains. However, although the previous voltage array V_z has Nc signals for the prediction calculation consideration period, if Nc>No, v_zNo+1, ..., v_zNc are not used.
[0123] First, the first current estimation calculation unit 19a inputs the output current i_out, the previous first output voltage v_z1, and the No-th output voltage before last v_zzNo which is the result of delaying the previous No-th output voltage v_zNo by one control period using the buffer 18. This first current estimation calculation unit 19a outputs the output current i_k-No+1 at time k-No+1.
[0124] Next, the second current estimation calculation unit 19b receives the output current i_k-No+1 at time k-No+1, the first previous output voltage v_z1, and the second previous output voltage v_z2, and outputs the output current i_k-No+2 at time k-No+2.
[0125] Next, the third current estimation calculation unit 19c receives the output current i_k-No+2 at time k-No+2, the second previous output voltage v_z2, and the third previous output voltage v_z3, and outputs the output current i_k-No+3 at time k-No+3.
[0126] This configuration is repeated, and finally, the No. current estimation calculation unit 19No receives the output current i_k-1 at time k-1, the No-1 previous output voltage v_zNo-1, and the No previous output voltage v_zNo to obtain the output current i_k at time k. The output current i_k at time k becomes the output of the current change estimation unit 9.
[0127] Using first to No. current estimation calculation units 19a to 19No that estimate the current one cycle ahead, an output current is estimated for each cycle of the previously determined output voltage. This calculation is repeated until time j+1 in equation (14) described later becomes time k, and an output current i_k at time k is obtained.
[0128] Considering a general (arbitrary) time j, the current estimation calculation unit 19 calculates the output current i_j+1 at time j+1 from the output current i_j at time j and the output voltage v_j at time j. However, in order to estimate the voltage during dead time, the output voltage v_j-1 from one cycle before is also required. Therefore, in addition to the voltage for that cycle, the voltage from one cycle before is also input to the current estimation calculation unit 19 for each cycle. For the first cycle, the voltage from one cycle before last, v_zzNo, is used, which is the output voltage v_zNo from the No cycle delayed once.
[0129] Fig. 11 shows the current estimation calculation unit 19. This shows the inside of the first to No. current estimation calculation units 19a to 19No in Fig. 10. As in Fig. 9, for generalization, the distinction between the first circuit and the second circuit is eliminated, and the time is also designated as an input / output name based on a general (arbitrary) time j.
[0130] A dead-time voltage estimating unit 20 receives an output current i_j at time j, an output voltage v_j-1 at time j-1, and an output voltage v_j at time j, and outputs an output voltage VH_d during the dead time.
[0131] The selector SW4 inputs the positive DC power supply voltage VH_in, 0, and the negative DC power supply voltage -VH_in. The selector SW4 also inputs the output voltage v_j at time j as the selection signal in4, and outputs the positive DC power supply voltage VH_in when the selection signal in4=1, outputs 0 when the selection signal in4=2, and outputs the negative DC power supply voltage -VH_in when the selection signal in4=3.
[0132] The multiplier 21 multiplies the output voltage VH_d during the dead time by Tdead / Tc. The multiplier 22 multiplies the output of the selector SW4 by (1-Tdead) / Tc. The adder 23 adds the output of the multiplier 21 and the output of the multiplier 22 together.
[0133] Multiplier 24 multiplies the output current i_j at time j by -R_load. Adder 25 adds the output of multiplier 24 to the output of adder 23. Multiplier 26 multiplies the output of adder 25 by Tc / L_load. The output of multiplier 26 becomes the current change amount Delta_i in one period. Adder 27 adds the output current i_j at time j to the current change amount Delta_i in one period and outputs the result as output current i_j+1 at time j+1.
[0134] This configuration estimates the current one cycle ahead. The formulas for calculating the current one cycle ahead of the H-bridge circuit connected to the electromagnet winding are (11) and (12). Here, L_load is the winding inductance [H], R_load is the winding resistance [Ω], Tc is the output voltage switching period [s], and VH_out is the average output voltage [V] that the H-bridge circuit provides between the load terminals during the output voltage switching period Tc.
[0135]
number
[0136]
number
[0137] The average output voltage VH_out is determined by equation (13), where VH_j is the output voltage of the H-bridge circuit corresponding to the state of the output voltage at time j, and VH_d is the output voltage during the dead time, where Tdead is the dead time [s].
[0138]
number
[0139] By substituting equations (12) and (13) into equation (11), we obtain equation (14).
[0140]
number
[0141] The block diagram in FIG. 11 performs the calculation of equation (14).
[0142] The output voltage VH_j at time j is determined by equation (15) based on the definition of the output voltage v_j at time j. In FIG. 11, the selector SW4 performs an operation corresponding to equation (15).
[0143]
number
[0144] Fig. 12 shows the dead-time voltage estimator 20. This shows the inside of the dead-time voltage estimator 20 in Fig. 11. As in Fig. 9, for the sake of generalization, the distinction between the first circuit and the second circuit is eliminated, and the time is also designated as an input / output name based on a general (arbitrary) time j.
[0145] Comparator GRT1 receives output current i_j at time j and dead time determination current i_dead, and outputs 1 if output current i_j at time j is greater than dead time determination current i_dead, and outputs 0 otherwise.
[0146] Comparator LES2 receives the output current i_j at time j and the dead time determination current -i_dead, and outputs 1 if the output current i_j at time j is smaller than the dead time determination current -i_dead, and outputs 0 otherwise.
[0147] The subtractor 28 subtracts the output voltage (state) v_j-1 at time j-1 from the output voltage (state) v_j at time j.
[0148] Comparator EQ1 receives the output of subtractor 28 and constant -1, and if the output of subtractor 28 is the same as constant -1, it outputs 1, and otherwise outputs 0. Comparator EQ2 receives the output of subtractor 28 and constant 1, and if the output of subtractor 28 is the same as constant 1, it outputs 1, and otherwise outputs 0.
[0149] The AND logic AND2 inputs the outputs of the comparators EQ1 and GRT1, and outputs 1 if both are 1, otherwise it outputs 0. The AND logic AND3 inputs the outputs of the comparators EQ2 and LES2, and outputs 1 if both are 1, otherwise it outputs 0.
[0150] The OR logic OR1 receives the outputs of the AND logic AND2 and AND3, and outputs 1 if at least one of them is 1, and outputs 0 if both are 0.
[0151] The selector SW5 inputs the output voltage (state) v_j at time j and the output voltage (state) v_j-1 at time j-1. The selector SW5 also inputs the output signal of the OR logic OR1 as a selection signal in5, and outputs the output voltage (state) v_j at time j if the selection signal in5=0, and outputs the output voltage (state) v_j-1 at time j-1 if the selection signal in5=1.
[0152] The selector SW6 inputs the positive DC power supply voltage VH_in, 0, and the negative DC power supply voltage -VH_in. The selector SW6 also inputs the output of the selector SW5 as a selection signal in6, and outputs the positive DC power supply voltage VH_in when the selection signal in6=1, outputs 0 when the selection signal in6=2, and outputs the negative DC power supply voltage -VH_in when the selection signal in6=3. The output of the selector SW6 becomes the output of the dead-time voltage estimation unit 20 as the dead-time output voltage VH_d.
[0153] This configuration estimates the output voltage during the dead time period.
[0154] In the first embodiment, however, the input on the upper side of the block of comparators EQ, NEQ, GRT, and LES in the figure is treated as the first input Lin1, and the input on the lower side is treated as the second input Lin2, and the output Lout is determined by the following equations (16) to (19). Equation (16) represents the comparator EQ, equation (17) represents the comparator NEQ, equation (18) represents the comparator GRT, and equation (19) represents the comparator LES.
[0155]
number
[0156]
number
[0157]
number
[0158]
number
[0159] In addition, the AND logic AND2, AND3 and the OR logic OR1 are treated the same as normal logic circuit symbols.
[0160] In addition, in the present embodiment 1, the logic circuit output is not treated as a logical variable, but as a variable that can be added or subtracted in the same way as an integer variable, as necessary. When performing processing involving different types, it is assumed that type conversion processing required for implementation is performed.
[0161] In the following, the output voltage during the dead time period of the first H-bridge circuit 3A will be considered, but the same applies to the second H-bridge circuit 3B.
[0162] Figure 13 shows the current path and output voltage during the dead time period. When the gate signal transitions from (1,0,0,1) to (0,1,0,1), a period (dead time) is provided in which both switching elements SA1 and SA2 are OFF to prevent a short circuit between switching elements SA1 and SA2. Here, the current path and output voltage during the dead time period (0,0,0,1) are shown.
[0163] Although switching element SA1 and switching element SA2 are OFF, at this time, current flows through the diode path of switching element SA2 from the direction of the first electromagnet winding current iA_out. As a result, current flows through the path that passes through switching element SA2 and switching element SA4, so the first H-bridge circuit output voltage VH_A becomes zero voltage.
[0164] For other combinations of transitions, consider the voltage during the dead time in the same way. Table 1 shows each gate signal transition and the voltage during the dead time.
[0165] [Table 1]
[0166] In the first embodiment, the first electromagnet winding current iA_out is positive, but the description has been given assuming that the first electromagnet winding current iA_out is negative as a general operation of the H-bridge circuit.
[0167] This is applied to the output voltage reference state of the first embodiment, and the table is simplified. Table 2 shows the voltages during each state transition and the dead time period.
[0168] [Table 2]
[0169] For state 2, the output voltage during the dead time period is the same for either gate signal, so the number of cases is reduced.
[0170] In Table 2, the output voltage after the dead time is listed next to the output voltage during the dead time. By comparing the voltage after the dead time, that is, the output voltage of the state after the transition, with the output voltage during the dead time, you can see when the presence or absence of dead time affects the output voltage and when it does not.
[0171] In the first row of Table 2, the output voltage during the dead time period is the same as the output voltage after the dead time ends, and the output voltage is not affected by the dead time. In such a case, it is acceptable to calculate the state after the transition even during the dead time period.
[0172] In the second row of Table 2, the output voltage during the dead time of the transition from state 2 to state 1 is different from the output voltage after the end of the dead time. In this case, the output voltage during the dead time is the same as the output voltage of the state before the transition. In such a case, the calculation is performed assuming that the state before the transition is in effect during the dead time.
[0173] If we consider the other rows in the same way, the output voltage during the dead time period can be determined by the following conditions. If iA_out>0, the state before the transition will be the state during the dead time if the state value is decreased by 1 from the previous value. Conversely, the state after the transition will be the state during the dead time if the state value is increased by 1 from the previous value. If iA_out<0, the state value will be in the state before the transition during the dead time if the transition increases the state value by 1 from the previous value. Conversely, the state will be in the state after the transition during the dead time if the transition decreases the state value by 1 from the previous value.
[0174] Taking this property into account, the output voltage during the dead time is estimated in Figure 12. The output voltage (state) v_j-1 at time j-1 is subtracted from the output voltage (state) v_j at time j, and the result is input to comparators EQ1 and EQ2.
[0175] Comparator EQ1 sees a match with -1, and in AND logic AND2 it determines if the first electromagnet winding current iA_out is positive and the dead time period is in the state before the transition. Comparator EQ2 sees a match with 1, and in AND logic AND3 it determines if the first electromagnet winding current iA_out is negative and the dead time period is in the state before the transition.
[0176] Here, the positive and negative currents are evaluated by comparators GRT1 and LES2. i_dead is the dead time determination current. Ideally, determination can be made when i_dead=0, but if a positive value of several percent to a few percent of the rated current (the maximum value of the current that can be continuously operated) is substituted for the dead time determination current i_dead, it is possible to prevent an increase in voltage error due to incorrect polarity determination near zero current.
[0177] When the AND logic AND2 or AND logic AND3 is 1, the state during the dead time period is the state before the transition. When the OR logic OR1 is 1, selector SW5 outputs the output voltage (state) v_j-1 at time j-1 before the transition, and when it is 0, selector SW5 outputs the output voltage (state) v_j at time j after the transition, thereby selecting an appropriate state as the output voltage during the dead time period.
[0178] The selector SW6 performs the operation corresponding to the equation (15) in the same manner as the selector SW4 in FIG. 11, and outputs an output voltage value corresponding to the state.
[0179] By the above operation, the output voltage during the dead time period can be estimated.
[0180] 14 shows the MPC evaluation calculation unit 10. This shows the inside of the MPC evaluation calculation unit 10 shown in FIG.
[0181] The input signals, namely, the first circuit command current iA_ref, the second circuit command current iB_ref, the first electromagnet winding current iA_k at time k, the second electromagnet winding current iB_k at time k, the first circuit previous voltage array VA_z, the second circuit previous voltage array VB_z, and the time k evaluation function value J_k with zero substituted, are collectively defined as the time k data array Dk.
[0182] The one-cycle prediction evaluation unit 291 inputs a time k data array Dk, and outputs a time k+1 pattern 1 data array D'k+1_1, a time k+1 pattern 2 data array D'k+1_2, . . . , a time k+1 pattern 9 data array D'k+1_9.
[0183] 9-input Ns-output sorting selection unit 301 inputs pattern 1 data array D'k+1_1 to pattern 9 data array D'k+1_9 at time k+1, and outputs the first place data array Dk+1_1 at time k+1, the second place data array Dk+1_2 at time k+1, ..., the Ns place data array Dk+1_Ns at time k+1.
[0184] The one-cycle prediction evaluation unit 2921 inputs the first place data array Dk+1_1 at time k+1, and outputs the previous first place pattern 1 data array D'k+2_11 at time k+2, ..., the previous first place pattern 9 data array D'k+2_19 at time k+2.
[0185] In addition, one cycle prediction evaluation units 2922 to 292Ns also input the second place data array Dk+1_2 at time k+1 to the Ns place data array Dk+1_Ns at time k+1, and output the previous second place pattern 1 data array D'k+2_21 at time k+2 to the previous Ns place pattern 9 data array D'k+2_Ns9 at time k+2.
[0186] The 9Ns-input, Ns-output sorting selection unit 302 inputs the previous 1st place pattern 1 data array D'k+2_11 to the previous Ns-place pattern 9 data array D'k+2_Ns9 at time k+2, and outputs the 1st place data array Dk+2_1 at time k+2, the second place data array Dk+2_2 at time k+2, ..., the Ns-place data array Dk+2_Ns at time k+2.
[0187] The same calculations are performed for the following data arrays, and finally, the data array D'k+Nc_11 that was the number 1 pattern 1 data array at time k+Nc last time to the data array D'k+Nc_Ns9 that was the Ns-th number pattern 9 data array at time k+Nc last time are input to the sorting selection unit 30Nc with 9Ns inputs and Ns outputs, and only the data array Dk+Nc_1 that was the number 1 data array at time k+Nc is output from the MPC evaluation calculation unit 10.
[0188] The MPC evaluation calculation unit 10 predicts and evaluates the current change for each pattern of the output voltage in one cycle, and selects the results with the best evaluation results. Every time a prediction evaluation is performed for one cycle, sorting and selection are performed, and Ns pieces of data with the best evaluation results are output. Ns is the number of evaluation results selected.
[0189] Then, prediction and evaluation are performed for the next cycle for each of the Ns outputs. The prediction and evaluation of the current change for each pattern of the output voltage for this cycle and the sorting and selection of the evaluation results are repeated until prediction and evaluation are performed for the next Nc cycles considered for prediction calculation, and the data related to the output voltage of the first H-bridge circuit and the second H-bridge circuit with the best evaluation results is obtained as the first-place data array Dk+Nc_1 at time k+Nc.
[0190] Regarding the time k data array Dk, in this embodiment 1, in order to simplify the notation of the MPC evaluation calculation, information required for prediction, evaluation, and sorting is summarized as the time k data array Dk. The time k data array Dk is defined by equation (20).
[0191]
number
[0192] The necessity of the sorting and selection unit 30 will be described. In MPC, it is ideal to predict and evaluate all combinations of output voltages up to the prediction calculation consideration period Nc ahead and confirm the best result. However, as will be described later, there are nine patterns of output voltage per period, so Nc This increases the number of combinations, and the computational load becomes unacceptable.
[0193] Therefore, the data array is sorted according to the evaluation function value for each calculation cycle, and only Ns combinations with relatively good evaluation functions are selected for evaluation in the next cycle. This reduces the number of combinations to {1+(Nc-1)Ns}9.
[0194] Fig. 15 shows the one-cycle prediction evaluation unit 29. This shows the inside of the one-cycle prediction evaluation units 291, 2921 to 292Ns, and 2931 to 293Ns in Fig. 14. For the input / output names, a general (arbitrary) time j is used as the reference for time for generalization, and the symbol indicating the sorting order of the data array is also set to a. Note that when calculating from time k to time k+1, there is no part corresponding to the sorting order a, but in this case, the sorting order a will be omitted as appropriate.
[0195] The a-th order data array Dj_a at time j is divided into the first circuit voltage array VA_j at time j, the second circuit voltage array VB_j at time j, the first circuit command current iA_ref, the second circuit command current iB_ref, the first circuit output current iA_j at time j, the second circuit output current iB_j at time j, and the evaluation function value J_j at time j, which are input to the pattern evaluation units 31a1 to 31a9.
[0196] In addition to the above-mentioned inputs, each of the pattern evaluation units 31a1 to 31a9 receives a signal related to the output voltage at time j+1. The pattern evaluation unit 31a1 located at the top of the drawing receives the first circuit output voltage (state) vA_j+1_1 at time j+1, pattern 1, and the second circuit output voltage (state) vB_j+1_1 at time j+1, pattern 1.
[0197] Here, in accordance with the definition of the pattern, 1 is substituted for the time j+1 pattern 1 first circuit output voltage (state) vA_j+1_1, and 1 is substituted for the time j+1 pattern 1 second circuit output voltage (state) vB_j+1_1. The output of each pattern evaluation unit 31a1 becomes the time j+1 previous a-th place pattern 1 data array D'j+1_a1.
[0198] The pattern evaluation unit 31a2 located second from the top in the figure receives the time j+1 pattern 2 first circuit output voltage (state) vA_j+1_2 and the time j+1 pattern 2 second circuit output voltage (state) vB_j+1_2. Here, in accordance with the pattern definition, 1 is substituted for the time j+1 pattern 2 first circuit output voltage (state) vA_j+1_2 and 2 is substituted for the time j+1 pattern 2 second circuit output voltage (state) vB_j+1_2. The output of each pattern evaluation unit 31a2 becomes the time j+1 previous a-th place pattern 2 data array D'j+1_a2.
[0199] The rest are configured similarly, and the first circuit output voltage (state) vA_j+1_9 at time j+1 pattern 9 and the second circuit output voltage (state) vB_j+1_9 at time j+1 pattern 9 are input to each pattern evaluation unit 31a9 located at the bottom of the figure. Here, in accordance with the pattern definition, 3 is substituted for the first circuit output voltage (state) vA_j+1_9 at time j+1 pattern 9 and 3 is substituted for the second circuit output voltage (state) vB_j+1_9 at time j+1 pattern 9. The output of each pattern evaluation unit 31a9 becomes the time j+1 previous a-th place pattern 9 data array D'j+1_a9.
[0200] The time j+1 previous a-th place pattern 1 data array D'j+1_a1 to the time j+1 previous a-th place pattern 9 data array D'j+1_a9 are output from prediction evaluation unit 29 for one period.
[0201] Here, data arrays D'j+1_a1, D'j+1_a2 to D'j+1_a9 of each pattern at time j+1 are calculated from the input time j ath place data array Dj_a. In addition to the signal obtained from the input time j ath place data array Dj_a, output voltage states such as time j+1 pattern 1 first circuit voltage vA_j+1_1 and time j+1 pattern 1 second circuit voltage vB_j+1_1 are input as voltages at time j+1 to each pattern evaluation unit 31a1 to 31a9.
[0202] The voltage value at time j+1 is different for each of the nine pattern evaluation units 31a1 to 31a9, which allows evaluation of nine possible patterns of output voltage and current for one cycle ahead. Table 3 shows the definitions of the voltage patterns and output voltage states.
[0203] [Table 3]
[0204] As shown in Table 3, there are nine patterns because there are three states for each of the first circuit output voltage vA and the second circuit output voltage vB. This Table 3 defines a pattern number for each combination of the first circuit output voltage vA and the second circuit output voltage vB. The input values for the time j+1 pattern x first circuit output voltage (state) vA_j+1_x and the time j+1 pattern x second circuit output voltage (state) vB_j+1_x in each of the nine pattern evaluation units 31a1 to 31a9 of the one-cycle prediction evaluation unit 29 are determined based on this Table 3.
[0205] FIG. 16 shows each pattern evaluation unit 31. This shows the inside of each pattern evaluation unit 31a1 to 31a9 in FIG. 15. For input / output names, the time is based on a general (arbitrary) time j for generalization. Also, the symbol indicating the sort order of the output data array (the symbol corresponding to a in FIG. 15) is omitted.
[0206] The current estimation calculation unit 32a receives the first circuit output current iA_j at time j, the first circuit output voltage (state) vA_j at time j, and the first circuit output voltage (state) vA_j+1 at time j+1, and outputs the first circuit output current iA_j+1 at time j+1.
[0207] The second circuit output current iB_j at time j, the second circuit output voltage (state) vB_j at time j, and the second circuit output voltage (state) vB_j+1 at time j+1 are input to the current estimation calculation unit 32b, and the second circuit output current iB_j+1 at time j+1 is output.
[0208] At this time, the first circuit output voltage (state) vA_j at time j and the second circuit output voltage (state) vB_j at time j are obtained by extracting the jth information from the first circuit voltage array VA_j at time j and the second circuit voltage array VB_j at time j in the jth selection units 33a and 33b.
[0209] Moreover, these current estimation calculation units 32a and 32b have the same configuration as the current estimation calculation unit 19 shown in FIG.
[0210] The evaluation calculation unit 34a1 receives the first circuit command current iA_ref and the first circuit output current iA_j+1 at time j+1, and outputs a first circuit current evaluation value JiA_j+1 at time j+1. The evaluation calculation unit 34a2 receives the first circuit output voltage (state) vA_j at time j and the first circuit output voltage (state) vA_j+1 at time j+1, and outputs a first circuit voltage evaluation value JvA_j+1 at time j+1.
[0211] The second circuit command current iB_ref and the second circuit output current iB_j+1 at time j+1 are input to the evaluation calculation unit 34b1, which outputs a second circuit current evaluation value JiB_j+1 at time j+1. Also, the second circuit output voltage (state) vB_j at time j and the second circuit output voltage (state) vB_j+1 at time j+1 are input to the evaluation calculation unit 34b2, which outputs a second circuit voltage evaluation value JvB_j+1 at time j+1.
[0212] The subtractor 35a subtracts the second circuit command current iB_ref from the first circuit command current iA_ref to output the command current difference iA-B_ref. The subtractor 35b subtracts the second circuit output current iB_j+1 at time j+1 from the first circuit output current iA_j+1 at time j+1 to output the output current difference iA-B_j+1 at time j+1.
[0213] The command current difference iA-B_ref and the output current difference iA-B_j+1 at time j+1 are input to evaluation calculation unit 34f, which outputs a current difference evaluation value JiF_j+1 at time j+1.
[0214] In adder 36, the first circuit current evaluation value JiA_j+1 at time j+1, the first circuit voltage evaluation value JvA_j+1 at time j+1, the second circuit current evaluation value JiB_j+1 at time j+1, the second circuit voltage evaluation value JvB_j+1 at time j+1, the current difference evaluation value JiF_j+1 at time j+1, and the evaluation function value J_j at time j are added together to obtain the evaluation function value J_j+1 at time j+1.
[0215] Voltage pattern recording unit 37a inputs the first circuit voltage array VA_j at time j and the first circuit output voltage vA_j+1 at time j+1, and outputs the first circuit voltage array VA_j+1 at time j+1. Voltage pattern recording unit 37b similarly inputs the second circuit voltage array VB_j at time j and the second circuit output voltage vB_j+1 at time j+1, and outputs the second circuit voltage array VB_j+1 at time j+1.
[0216] The signals of the first circuit voltage array VA_j+1 at time j+1, the second circuit voltage array VB_j+1 at time j+1, the first circuit command current iA_ref, the second circuit command current iB_ref, the first circuit output current iA_j+1 at time j+1, the second circuit output current iB_j+1 at time j+1, and the evaluation function value J_j+1 at time j+1 are compiled and output from each pattern evaluation unit 31 as a data array D'j+1 at time j+1.
[0217] Here, the current one cycle ahead is estimated, and the evaluation function values of the current and voltage are calculated. The evaluation function is output together with the voltage value one cycle ahead and the estimated current value as a data array D'j+1 at time j+1.
[0218] The current estimation calculation unit 32a of each pattern evaluation unit 31 has the same configuration as the current estimation calculation unit 19 present in the current change estimation unit 9 described above. Therefore, the first circuit output current iA_j+1 at time j+1 is estimated from the first circuit output current iA_j at time j and the first circuit output voltage vA_j+1 at time j+1. At this time, in order to estimate the output voltage during the dead time period, the first circuit output voltage vA_j at time j is extracted from the first circuit voltage array VA_j at time j and input to the current estimation calculation unit 32a. The same is true for the current estimation calculation unit 32b.
[0219] The first circuit output voltage vA_j+1 at time j+1 and the second circuit output voltage vB_j+1 at time j+1 are input to voltage pattern recording units 37a, 37b along with the first circuit voltage array VA_j at time j and the second circuit voltage array VB_j at time j, respectively, and the first circuit output voltage vA_j+1 at time j+1 and the second circuit output voltage vB_j+1 at time j+1 are substituted into the j+1th position of the first circuit voltage array VA_j at time j and the second circuit voltage array VB_j at time j, and the first circuit voltage array VA_j+1 at time j+1 and the second circuit voltage array VB_j+1 at time j+1 are output.
[0220] In the evaluation calculation unit, the first circuit command current iA_ref and the first circuit output current iA_j+1 at time j+1 are input to the evaluation calculation unit 34a1, an evaluation is performed on the current value, and the first circuit current evaluation value JiA_j+1 at time j+1 is output.
[0221] The first circuit output voltage vA_j at time j and the first circuit output voltage vA_j+1 at time j+1 are input to the evaluation calculation unit 34a2, an evaluation is performed on the output voltage, and a first circuit voltage evaluation value JvA_j+1 at time j+1 is output.
[0222] For the second H-bridge circuit 3B, the evaluation calculation units 34b1 and 34b2 perform similar calculations to obtain a second circuit current evaluation value JiB_j+1 at time j+1 and a second circuit voltage evaluation value JvB_j+1 at time j+1.
[0223] The command current difference iA-B_ref and the estimated output current difference iA-B_j+1 at time j+1 are input to the evaluation calculation unit 34f. Here, the current difference is evaluated as an evaluation specific to the MPC of the magnetic levitation system, and a current difference evaluation value JiF_j+1 at time j+1 is obtained.
[0224] By adding together the first circuit current evaluation value JiA_j+1 at time j+1, the first circuit voltage evaluation value JvA_j+1 at time j+1, the second circuit current evaluation value JiB_j+1 at time j+1, the second circuit voltage evaluation value JvB_j+1 at time j+1, the current difference evaluation value JiF_j+1 at time j+1, and the evaluation function value J_j at time j, the evaluation function value J_j+1 at time j+1 is obtained.
[0225] This means that we have obtained the cumulative evaluation results for current and voltage up to time j+1. However, although many results are added up here, the process is performed in such a way that an evaluation value close to the maximum value does not suddenly decrease to the minimum value due to an overflow or the like.
[0226] FIG. 17 shows the evaluation calculation units 341, 342, and 34f of the first embodiment. This shows the inside of the evaluation calculation units 34a1, 34b1, evaluation calculation units 34a2, 34b2, and evaluation calculation unit 34f in FIG. 16. For input / output names, the time is based on a general (arbitrary) time j for generalization. Also, the evaluation calculation units 341 and 342 are described without distinguishing between the first circuit and the second circuit.
[0227] In evaluation calculation unit 341, subtractor 38 subtracts command current i_ref at time j from output current i_j at time j. Two-input multiplier 39 outputs the square of the output of subtractor 38. Multiplier 40 multiplies the output of two-input multiplier 39 by current evaluation gain G_Ji.
[0228] The absolute value calculation unit 41 outputs the absolute value of the output of the subtractor 38. The comparator GRT2 receives the output of the absolute value calculation unit 41 and the current deviation allowable value i_abs_ref, and outputs 1 if the output of the absolute value calculation unit 41 is greater than the current deviation allowable value i_abs_ref, and outputs 0 otherwise.
[0229] The selector SW7 receives a constant 0 and a deviation penalty gain G_Jiabs. The selector SW7 also receives the output of the comparator GRT2 as a selection signal in7, and outputs the constant 0 if the selection signal in7=0, and outputs the deviation penalty gain G_Jiabs if the selection signal in7=1.
[0230] The adder 42 adds the output of the multiplier 40 and the output of the selector SW7, and outputs a current evaluation value Ji_j at time j.
[0231] In the evaluation calculation section 342, the subtractor 43 subtracts the output voltage (state) v_j+1 at time j+1 from the output voltage (state) v_j at time j. The absolute value calculation section 44 outputs the absolute value of the output of the subtractor 43.
[0232] Comparator GRT3 receives the output of absolute value calculation unit 44 and constant 1, and outputs 1 if the output of absolute value calculation unit 44 is greater than constant 1, and otherwise outputs constant 0. Comparator GRT4 receives the output of absolute value calculation unit 44 and constant 0, and outputs 1 if the output of absolute value calculation unit 44 is greater than constant 0, and otherwise outputs constant 0. Adder 45 adds the output of comparator GRT3 and the output of comparator GRT4.
[0233] Selector SW8 inputs a constant 0, a voltage evaluation gain G_Jv, and a prohibited transition penalty gain G_JNG. Selector SW8 also inputs the output of adder 45 as selection signal in8, and outputs a constant 0 when selection signal in8=0, outputs a voltage evaluation gain G_Jv when selection signal in8=1, and outputs a prohibited transition penalty gain G_JNG when selection signal in8=2. This output is the voltage evaluation value Jv_j at time j. The above is evaluation calculation unit 342.
[0234] In evaluation calculation unit 34f, subtractor 46 subtracts command current difference iA-B_ref at time j from output current difference iA-B_j at time j. Two-input multiplier 47 outputs the square of the output of subtractor 46. Multiplier 48 multiplies the output of two-input multiplier 47 by current difference evaluation gain G_JiF.
[0235] The absolute value calculation unit 49 outputs the absolute value of the output of the subtractor 46. The comparator GRT5 receives the output of the absolute value calculation unit 49 and the current difference deviation allowable value iF_abs_ref, and outputs 1 if the output of the absolute value calculation unit 49 is greater than the current difference deviation allowable value iF_abs_ref, and outputs 0 otherwise.
[0236] The selector SW9 receives as input a constant 0 and a differential deviation penalty gain G_JiFabs. The selector SW9 also receives as input the output of the comparator GRT5 as a selection signal in9, and outputs the constant 0 when the selection signal in9=0, and outputs the differential deviation penalty gain G_JiFabs when the selection signal in9=1.
[0237] The adder 50 adds the output of the multiplier 48 and the output of the selector SW9, and outputs a current difference evaluation value JiF_j at time j.
[0238] Here, the evaluation calculation of the current value and the voltage value is shown. The evaluation function is defined such that the larger the evaluation result value, the worse the evaluation result.
[0239] The evaluation calculation unit 341 evaluates the current by calculation of equations (21) and (22), where SW7out is the output value of the selector SW7.
[0240]
number
[0241]
number
[0242] The term of the current evaluation gain G_Ji in equation (21) increases exponentially as the output current value deviates from the command. As a result, the evaluation function value becomes smaller as the estimated current value approaches the command, and the output voltage where the current approaches the command is output preferentially.
[0243] In addition, the deviation penalty gain G_Jiabs term becomes the deviation penalty gain G_Jiabs when the difference between the output current value and the command exceeds the current deviation allowable value i_abs_ref, and becomes zero when it does not exceed the current deviation allowable value i_abs_ref. This term prohibits the output current from momentarily deviating from the command value, even though the current matches the command closely overall.
[0244] The evaluation calculation unit 342 evaluates the output voltage by the calculation of equation (23).
[0245]
number
[0246] In equation (23), zero is output when the output voltage state does not change, a voltage evaluation gain G_Jv is output when the state changes by 1, i.e., when there is a normal transition, and a prohibited transition penalty gain G_JNG is output when the state changes by 2, i.e., when there is an abnormal transition.
[0247] The voltage evaluation gain G_Jv prevents an increase in semiconductor loss by changing the output voltage, i.e., by making the evaluation worse the more the switching element switches. The prohibited transition penalty gain G_JNG is a gain for preventing prohibited state transitions. In order to ensure that the evaluation function value becomes the worst evaluation result during prohibited transitions, a positive maximum value should be substituted for the prohibited transition penalty gain G_JNG after taking measures against overflow in the subsequent stage.
[0248] The presence or absence of state transitions and abnormal transitions are observed by comparators GRT3 and GRT4. If there is a transition, comparator GRT4 becomes 1, and if there is an abnormal transition of 2 or more, comparator GRT3 also becomes 1, so the operation of equation (23) can be realized by inputting the added value of the two outputs to selection signal in8. However, as mentioned above, the output of comparator GRT is assumed to be an integer type that can be added.
[0249] In conventional MPC, it was sufficient to take into consideration the evaluation contents of the evaluation calculation units 341 and 342 and, if necessary, the motor speed command agreement in the case of a motor drive, the capacitor voltage in the case of a chopper, etc. Also, the contents of the evaluation calculation units 341 and 342 are completely independent evaluations for the first H-bridge circuit 3A and the second H-bridge circuit 3B, and no problem occurred even if the first H-bridge circuit 3A and the second H-bridge circuit 3B were controlled separately.
[0250] However, in a magnetic levitation system, the evaluation of the current difference becomes important because the current difference between the two circuits (converters) becomes the attractive force acting on the levitation body 1. The evaluation calculation unit 34f indicates an evaluation of the current difference suitable for the magnetic levitation system.
[0251] The evaluation calculation unit 34f evaluates the current difference by calculation of equations (24) and (25). Here, SW9out is the output value of the selector SW9. Here, iF_j is the same as iA-B_j, and iF_ref is the same as iA-B_ref.
[0252]
number
[0253]
number
[0254] The term of the current difference evaluation gain G_JiF in equation (24) increases exponentially as the output current difference deviates from the command current difference. As a result, the evaluation function value becomes smaller as the output current difference approaches the command current difference, and the output voltage where the output current difference approaches the command current difference is output preferentially.
[0255] In addition, the term of difference deviation penalty gain G_JiFabs becomes the difference deviation penalty gain G_JiFabs when the difference between the output current difference and the command current difference exceeds the current difference deviation allowable value iF_abs_ref, and becomes zero when it does not exceed the current difference deviation allowable value iF_abs_ref. This term prohibits an output current that momentarily deviates from the command current difference, although the output current difference is highly consistent with the command current difference overall.
[0256] In this way, by adding the degree of agreement with the command current difference not only for the output currents of the first and second circuits but also for the output current difference to the evaluation function, it is possible to reduce pulsation in the attractive force.
[0257] The evaluation method described above with reference to FIGS. 16 and 17 can also be applied to a magnetic levitation system that does not use an H-bridge circuit.
[0258] Fig. 18 shows the sorting selection unit 30 of the present embodiment 1. This shows the inside of the sorting selection units 301, 302, 303, ..., 30Nc in Fig. 14. The number of input signals Nin is used to generalize the number of inputs.
[0259] The evaluation value comparison unit C11 receives the (Nin-1)th data array D'j_Nin-1 at time j (before sorting) and the (Nin)th data array D'j_Nin at time j (before sorting).
[0260] A first output of the evaluation value comparison unit C11 is input to the evaluation value comparison unit C12 in the next higher stage, and a second output of the evaluation value comparison unit C11 is input to the evaluation value comparison unit C21. In addition, the (Nin-2)th data array D'j_Nin-2 at time j (before sorting) is input to the evaluation value comparison unit C12.
[0261] This configuration is connected in multiple stages to evaluation value comparison units C11, C12, C13, . . . , and finally, the first output of evaluation value comparison unit C1Nin-1 becomes the first-place data array Dj_1 at time j.
[0262] Similarly, the second output is also multi-staged with evaluation value comparison units C21, C22, C23, . . . , and the first output of evaluation value comparison unit C2Nin-2 becomes the second-place data array Dj_2 at time j.
[0263] The output signals are determined using this configuration, but when the Ns-th data array Dj_Ns at time j, which is the first output of evaluation value comparison unit CNsNin-Ns, is obtained, all output signals are complete. No evaluation value comparison units are provided thereafter. In other words, evaluation value comparison units CNs+11, CNs+12, ..., CNin1 do not exist. Also, since there is no next evaluation value comparison unit, the second outputs of evaluation value comparison units CNs1, CNs2, ..., CNsNin-Ns are not used.
[0264] The sorting uses the concept of a sorting network that does not require repeated calculations and is suitable for implementation in an FPGA (Field Programmable Gate Array). In this embodiment, all input signals are sorted and selected at once by the sorting network.
[0265] An example of the operation of a 5-input sorting network is shown in Figure 19. In the sorting network, sorting is performed by repeatedly switching signals by evaluation value comparison units C11-C14, C21-C23, C31-C32, and C41. Evaluation value comparison units C11-C14, C21-C23, C31-C32, and C41 compare the magnitude of two input signals and change the correspondence of the signals to the first and second outputs based on the comparison results. Here, the smaller of the inputs to evaluation value comparison units C11-C14, C21-C23, C31-C32, and C41 is the first output (the output at the top of the figure), and the larger is the second output (the output at the bottom of the figure).
[0266] For example, the fourth input, 1, is the smallest value among the five inputs, so it is output as the first output each time it is input to the evaluation value comparison unit, and ultimately becomes the top signal. The other signals are compared in the same way, and it can be seen that the five signals that were randomly arranged in terms of size at the time of input are sorted from smallest to largest at the output side.
[0267] Similarly, in FIG. 18, evaluation value comparison units C11, C12, etc. compare evaluation function values from data array information and repeatedly replace signals to sort the signals.
[0268] As for the sorting results, only the top Ns signals in the figure are output, and the rest are unused. This allows only the top Ns evaluation function values to be selected and passed to the subsequent stage. In this case, since the unused signals from the Ns+1th signal onwards do not need to be sorted, there are no CNs+11 and onwards.
[0269] Figure 20 shows the evaluation value comparison unit C. This shows the inside of the evaluation value comparison units C11 to CNSNin-Ns in Figure 18. For generalization, the time is based on a general (arbitrary) time j. Also, the symbol indicating the sort order of the output data array (the symbol corresponding to a in Figure 15) is omitted.
[0270] The J_j extraction unit 51 extracts the time j evaluation function value J_j_in1 of the first input from the internal information of the time j data array D'j_in1 of the first input. The J_j extraction unit 52 extracts the time j evaluation function value J_j_in2 of the second input from the internal information of the time j data array D'j_in2 of the second input. The comparator GRT6 receives the time j evaluation function value J_j_in1 of the first input and the time j evaluation function value J_j_in2 of the second input, and outputs 1 if the time j evaluation function value J_j_in1 of the first input is greater than the time j evaluation function value J_j_in2 of the second input, and outputs 0 otherwise.
[0271] The selector SW10 receives the time j data array D'j_in1 as a first input and the time j data array D'j_in2 as a second input. The selector SW10 also receives the output signal of the comparator GRT6 as a selection signal in10, and outputs the time j data array D'j_in1 as the first input when the selection signal in10=0, and outputs the time j data array D'j_in2 as the second input when the selection signal in10=1.
[0272] The selector SW11 receives the second input time j data array D'j_in2 and the first input time j data array D'j_in1. The selector SW11 also receives the output signal of the comparator GRT6 as a selection signal in11, and outputs the second input time j data array D'j_in2 when the selection signal in11=0, and outputs the first input time j data array D'j_in1 when the selection signal in11=1.
[0273] The output of the selector SW10 becomes the time j data array D'j_out1 of the first output, and the output of the selector SW11 becomes the time j data array D'j_out2 of the second output.
[0274] The evaluation value comparison unit C extracts information about the evaluation function values J_j from the input data array, and performs processing to switch output signals depending on the magnitude relationship.
[0275] The magnitude relationship is evaluated by comparator GRT6, and the output signals of selectors SW10 and SW11 are determined based on the comparison results. Since it is defined that the smaller the evaluation function value, the better the result, the data array with the smallest evaluation function value J_j is configured to be the first output.
[0276] The selector SW10 corresponds to the time j data array D'j_out1 of the first output, and the selector SW11 corresponds to the time j data array D'j_out2 of the second output, and the input / output relationship is expressed by equations (26) and (27).
[0277]
number
[0278]
number
[0279] The operation of the evaluation value comparison unit C has been described above.
[0280] As described above, by performing control based on Figs. 1 to 20, it is possible to perform MPC that is suitable for controlling the H-bridge circuit and for controlling the attractive force of the magnetic levitation system.
[0281] [Embodiment 2] Fig. 21 shows the sorting selection unit 30 of the present embodiment 2. This shows the inside of the sorting selection units 301, 302, 303, ..., 30Nc in Fig. 14. For generalization, a general (arbitrary) time j is used as the reference time.
[0282] The 9-input, 3-output sorting selection unit 30 here has a configuration in which Nin=9, Ns=3 in FIG.
[0283] The nine signals of the previous first place pattern 1 data array D'j_11 to the previous first place pattern 9 data array D'j_19 at time j are input to a 9-input, 3-output sorting selection unit 53a. The nine signals of the previous second place pattern 1 data array D'j_21 to the previous second place pattern 9 data array D'j_29 at time j are input to a 9-input, 3-output sorting selection unit 53b. The nine signals of the previous third place pattern 1 data array D'j_31 to the previous third place pattern 9 data array D'j_39 at time j are input to a 9-input, 3-output sorting selection unit 53c.
[0284] A total of nine output signals from the three 9-input, 3-output sorting selection units 53a to 53c are input to a 9-input, 3-output sorting selection unit 53d, which outputs the first place data array Dj_1 at time j to the third place data array Dj_3 at time j.
[0285] In the first embodiment, all input signals are sorted at once by the sorting network in the sorting selection unit 30. However, sorting does not necessarily have to be performed in one sorting network, and may be performed in multiple stages.
[0286] In a multi-input, multi-output sorting network, the implementation becomes complicated due to the increase in the number of evaluation value comparison units per module, and the number of possible combinations of inputs becomes enormous, making debugging a burden. Therefore, the burden of implementation can be reduced by creating and combining sorting networks in as small units as possible.
[0287] When making multiple stages, care must be taken to ensure that the number of outputs of each sorting network is equal to or greater than the evaluation result selection number Ns. If there is a sorting network with outputs less than the evaluation result selection number Ns, the 1st to Nsth ranked signals may be present in the truncated outputs, making it impossible to perform sorting correctly.
[0288] In the second embodiment, the sorting network is multi-staged, the input signal is divided into a plurality of groups, and sorted and selected by the plurality of sorting networks, and the selected signal is sorted and selected by the sorting network of the next stage.
[0289] Fig. 21 shows the sorting selection unit 30 of the second embodiment. This is an example of multi-stage sorting when Ns = 3. In this configuration, four 9-input, 3-output sorting selection units 53a to 53d are combined for 27 inputs and 3 outputs to achieve 27 inputs and 3 outputs sorting. In this way, sorting may be performed by multi-stage small-scale sorting networks for a multi-input, multi-output sorting network.
[0290] As described above, the second embodiment provides the same advantageous effects as those of the first embodiment. Furthermore, the second embodiment can reduce the burden of implementation.
[0291] [Embodiment 3] Fig. 22 shows the evaluation calculation unit of the third embodiment. In the third embodiment, an evaluation calculation unit 341' is applied in place of the evaluation calculation units 34a1 and 34b1 in Fig. 16, and an evaluation calculation unit 34f' is applied in place of the evaluation calculation unit 34f. The evaluation calculation units 34a2 and 34b2 are similar to those in the first embodiment, and therefore illustration and description thereof will be omitted.
[0292] For the sake of generalization, the input and output names are based on a general (arbitrary) time j. Also, the names are written without distinction between the first and second circuits.
[0293] In evaluation calculation unit 341', subtractor 54 subtracts command current i_ref at time j from output current i_j at time j. Absolute value calculation unit 55 calculates the absolute value of the output of subtractor 54. Multiplier 56 multiplies the output of absolute value calculation unit 55 by current evaluation gain G_Ji.
[0294] Comparator GRT7 receives the output of absolute value calculation unit 55 and current deviation allowable value i_abs_ref, and outputs 1 if the output of absolute value calculation unit 55 is greater than current deviation allowable value i_abs_ref, and outputs 0 otherwise.
[0295] The selector SW12 receives a constant 0 and a deviation penalty gain G_Jiabs. The selector SW12 also receives the output of the comparator GRT7 as a selection signal in12, and outputs the constant 0 when the selection signal in12=0, and outputs the deviation penalty gain G_Jiabs when the selection signal in12=1.
[0296] The adder 57 adds the output of the multiplier 56 and the output of the selector SW12, and outputs the result as a current evaluation value Ji_j at time j. The above is the evaluation calculation unit 341'.
[0297] In evaluation calculation unit 34f', subtractor 58 subtracts command current difference iA-B_ref at time j from output current difference iA-B_j at time j. Absolute value calculation unit 59 calculates the absolute value of the output of subtractor 58. Multiplier 60 multiplies the output of absolute value calculation unit 59 by current difference evaluation gain G_JiF.
[0298] Comparator GRT8 receives the output of absolute value calculation unit 59 and current difference deviation allowable value iF_abs_ref, and outputs 1 if the output of absolute value calculation unit 59 is greater than current difference deviation allowable value iF_abs_ref, and outputs 0 otherwise.
[0299] The selector SW13 receives a constant 0 and a differential deviation penalty gain G_JiFabs. The selector SW13 also receives the output of the comparator GRT8 as a selection signal in13, and outputs the constant 0 when the selection signal in13=0, and outputs the differential deviation penalty gain G_JiFabs when the selection signal in13=1.
[0300] The adder 61 adds the output of the multiplier 60 to the output of the selector SW13 and outputs the result as a current difference evaluation value JiF_j at time j. The above is the evaluation calculation unit 34f'.
[0301] In the evaluation calculation unit 341 and the evaluation calculation unit 34f in the first and second embodiments, the square of the current difference is evaluated.
[0302] However, the absolute value of the current difference may be evaluated, which makes it possible to reduce the number of multipliers used, resulting in an advantageous implementation.
[0303] 22 shows the evaluation calculation unit of the present embodiment 3. The evaluation calculation unit 341' can be used in place of the evaluation calculation unit 341, and the evaluation calculation unit 34f' can be used in place of the evaluation calculation unit 34f.
[0304] In the evaluation calculation unit 341', the absolute value of the difference between the command current and the output current is calculated, and then multiplied by the current evaluation gain G_Ji. The other parts are the same as those of the evaluation calculation unit 341, and the current evaluation value Ji_j is expressed by equations (28) and (29). However, SW12out is the output value of the selector SW12.
[0305]
number
[0306]
number
[0307] In evaluation calculation unit 34f', the absolute value of the deviation between the command current difference and the output current difference is taken, and then multiplied by the current difference evaluation gain G_JiF. The other parts are the same as in evaluation calculation unit 34f, and the current difference evaluation value JiF_j is expressed by equations (30) and (31). Here, iF_j is the same as iA-B_j, and iF_ref is the same as iA-B_ref.
[0308]
number
[0309]
number
[0310] As described above, according to the third embodiment, it is possible to obtain the same advantageous effects as those of the first and second embodiments. Moreover, the third embodiment can reduce the number of multipliers used, which is advantageous in terms of implementation.
[0311] [effect] Figure 23 shows the simulation results of the first embodiment. The left side of Figure 23 is a comparative example, and the right side is the result of the first embodiment. Feedback control was performed by MPC, and the waveforms of the first electromagnet winding current iA_out, the second electromagnet winding current iB_out, the current difference iA_out-iB_out, the command current difference iAref-iBref, and the switching frequencies fswA and fswB of the first H-bridge circuit 3A and the second H-bridge circuit 3B were recorded. For the switching frequencies fswA and fswB, a moving average was taken of the number of switching times in the most recent 10 ms period, and the results were converted to a period of 1 s. Table 4 shows the constant settings for the simulation.
[0312] [Table 4]
[0313] In this simulation, the current control results of the comparative example and the technology of embodiment 1 are compared. However, the comparative example is based on the premise that MPC can be applied to an H-bridge circuit in the prior art, and expresses a situation in which the currents of two circuits (converters) are individually controlled by MPC using gain adjustment of the technology of embodiment 1. In the technology of embodiment 1, the current difference between the two circuits is included in the evaluation function, and the mutual relationship between the two circuits is taken into consideration.
[0314] In the comparative example, the current difference evaluation gain G_JiF and the difference deviation penalty gain G_JiFabs in the evaluation function were set to zero, making it equivalent to the prior art corresponding to an H-bridge circuit. In other words, the current difference between the two circuits was not taken into consideration, and each current was controlled so that it matched the command to a high degree. As a result, although the pulsation of the first electromagnet winding current iA_out and the second electromagnet winding current iB_out was suppressed, the pulsation of the current difference was larger than the pulsation of each winding current, especially at the timing when the pulsations of the two circuits overlap.
[0315] In the technology of the first embodiment, the current difference between the two circuits is taken into consideration in the evaluation function. As a result, even if the pulsation of the first electromagnet winding current iA_out and the second electromagnet winding current iB_out is larger than that of the comparative example, the pulsation of the current difference is kept small.
[0316] In the linear region of the electromagnet, the resultant attractive force acting on the levitation body is proportional to the difference in winding currents, so the force pulsation on the levitation body can be reduced by reducing the pulsation of the current difference.
[0317] Looking at the switching frequencies of both methods, the switching frequencies fswA and fswB are both smaller in embodiment 1. By suppressing the switching frequencies fswA and fswB, the technology of embodiment 1 is expected to reduce the switching loss of the H-bridge circuit.
[0318] Therefore, by using the evaluation function including the evaluation of the current difference in the technology of embodiment 1, it was possible to reduce the pulsation of the current difference, that is, the force pulsation in the levitation body, compared to the comparative example while keeping the switching frequency low.
[0319] In this simulation, the comparative example is expressed by adjusting some of the gains of the technology of embodiment 1. In other words, the technology of embodiment 1 includes the configuration of the comparative example. By adjusting the gain in the technical configuration of embodiment 1, it is possible to control with characteristics equivalent to the comparative example or intermediate between the comparative example and the technology of embodiment 1. In the case where the pulsation of the attractive force is highly evaluated and the case where the individual current pulsations of the electromagnet windings are highly evaluated, the comparative example, which is equivalent to the prior art corresponding to an H-bridge circuit, can only handle the latter, whereas the technology of embodiment 1 can handle both. As described above, the technology of embodiment 1 can be said to be advantageous in terms of adjustability as well.
[0320] Although the present invention has been described in detail above only with respect to the specific examples, it will be apparent to those skilled in the art that various modifications and variations are possible within the scope of the technical concept of the present invention, and it is natural that such modifications and variations fall within the scope of the claims. [Explanation of symbols]
[0321] 1...Floating body 2A, 2B...1st, 2nd electromagnet 3A, 3B...1st, 2nd H bridge circuit (1st, 2nd circuit) 4…DC power supply 5. MPC 6...Multiplier 7...Adder 8…Subtractor 9A, 9b...Current change estimation section 10…MPC evaluation calculation section 11...Switching pattern outlet 12A, 12B...Buffer 13…Current output selection section 14A, 14B...Gate determination section
Claims
1. First and second electromagnets are provided at positions facing each other with respect to the levitation body; a first circuit for controlling a first electromagnet winding current and a second circuit for controlling a second electromagnet winding current; A DC power source that serves as a power source for the first and second circuits; an MPC that evaluates a first circuit current evaluation value, a first circuit voltage evaluation value, a second circuit current evaluation value, a second circuit voltage evaluation value, and a current difference evaluation value between the first circuit and the second circuit in an output voltage from the next period onward, taking into account models of the first circuit and the winding of the first electromagnet and the second circuit and the winding of the second electromagnet, and outputs a gate signal of the first circuit and a gate signal of the second circuit corresponding to the output voltage with the best evaluation result; An actively controlled magnetic levitation system comprising:
2. the first circuit is a first H-bridge circuit, and the second circuit is a second H-bridge circuit, When the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit are positive DC power supply voltages, the state is 1. When the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit are 0, the state is 2. When the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit are negative DC power supply voltages, the state is 3. a pattern of the gate signal of the first H-bridge circuit and a pattern of the gate signal of the second H-bridge circuit are one in the case of state 1, two in the case of state 2, and one in the case of state 3; 2. The actively controlled magnetic levitation system according to claim 1, wherein a transition from state 1 to state 3 and from state 3 to state 1 are prohibited, a pattern of the gate signal of the first H-bridge circuit and a pattern of the gate signal of the second H-bridge circuit are switched when transitioning from state 1 or state 3 to state 2 and when transitioning from state 2 to state 1 or state 3, the pattern of the gate signal is not switched while in state 2, and a pattern of the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit used in state 2 is a pattern different from the pattern of the previous state 2.
3. the first circuit is a first H-bridge circuit, and the second circuit is a second H-bridge circuit, When the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit are positive DC power supply voltages, the state is 3. When the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit are 0, the state is 2. When the output voltage of the first H-bridge circuit and the output voltage of the second H-bridge circuit are negative DC power supply voltages, the state is 1. a pattern of the gate signal of the first H-bridge circuit and a pattern of the gate signal of the second H-bridge circuit are one in the case of state 1, two in the case of state 2, and one in the case of state 3; 2. The actively controlled magnetic levitation system according to claim 1, wherein a transition from state 1 to state 3 and from state 3 to state 1 are prohibited, a pattern of the gate signal of the first H-bridge circuit and a pattern of the gate signal of the second H-bridge circuit are switched when transitioning from state 1 or state 3 to state 2 and when transitioning from state 2 to state 1 or state 3, the pattern of the gate signal is not switched while in state 2, and a pattern of the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit used in state 2 is a pattern different from the pattern of the previous state 2.
4. The MPC is Calculating the first circuit current evaluation value and the second circuit current evaluation value by equations (21) and (22), Calculating the first circuit voltage evaluation value and the second circuit voltage evaluation value by equation (23); 3. The active control type magnetic levitation system according to claim 2, wherein the current difference evaluation value is calculated by using equations (24) and (25). ##EQU00021## [0022] [0023] [0024] [0025] Ji_j: first circuit current evaluation value, second circuit current evaluation value at any time j G_Ji: Current evaluation gain i_j: output current at any time j i_ref: Command current G_Jiabs: deviation penalty gain i_abs_ref: current deviation tolerance Jv_j: first circuit voltage evaluation value, second circuit voltage evaluation value at any time j v_j: state of the output voltage at any time j v_j+1: state of the output voltage at time j+1 G_Jv: Voltage evaluation gain G_JNG: Forbidden transition penalty gain JiF_j: current difference evaluation value G_JiF: Current difference evaluation gain iA-B_j (=iF_j): Output current difference at any time j iA-B_ref (=iF_ref): Command current difference G_JiFabs: Differential deviation penalty gain iF_abs_ref: current difference deviation tolerance
5. The MPC is Calculating the first circuit current evaluation value and the second circuit current evaluation value by equations (28) and (29), Calculating the first circuit voltage evaluation value and the second circuit voltage evaluation value by equation (23); 3. The active control type magnetic levitation system according to claim 2, wherein the current difference evaluation value is calculated by using equations (30) and (31). [0028] [0029] [0023] [0030] [0031] Ji_j: first circuit current evaluation value, second circuit current evaluation value at any time j G_Ji: Current evaluation gain i_j: output current at any time j i_ref: Command current G_Jiabs: deviation penalty gain i_abs_ref: current deviation tolerance Jv_j: first circuit voltage evaluation value, second circuit voltage evaluation value at any time j v_j: state of the output voltage at any time j v_j+1: state of the output voltage at time j+1 G_Jv: Voltage evaluation gain G_JNG: Forbidden transition penalty gain JiF_j: current difference evaluation value G_JiF: Current difference evaluation gain iA-B_j (=iF_j): Output current difference at any time j iA-B_ref (=iF_ref): Command current difference G_JiFabs: Differential deviation penalty gain iF_abs_ref: current difference deviation tolerance
6. The MPC is a current change estimation unit that estimates a first electromagnet winding current at a currently determined output start time and a second electromagnet winding current at the currently determined output start time; an MPC evaluation calculation unit that outputs data on output voltages of the first H-bridge circuit and the second H-bridge circuit having the best evaluation results from the currently determined output start time to the prediction calculation consideration period, based on the first electromagnet winding current at the currently determined output start time, the second electromagnet winding current at the currently determined output start time, the first circuit command current of equation (7), and the second circuit command current of equation (8); a gate determination unit that outputs the gate signal of the first H-bridge circuit and the gate signal of the second H-bridge circuit based on an output voltage of the first H-bridge circuit and an output voltage of the second H-bridge circuit; 3. The actively controlled magnetic levitation system according to claim 2, further comprising: [0070] [0080] iA_ref: 1st circuit command current iB_ref: Second circuit command current ibias: command bias current iFref: command attraction force control current
7. The MPC evaluation calculation unit is 7. The actively controlled magnetic levitation system according to claim 6, further comprising: predicting and evaluating a current change for each pattern of the output voltage in one cycle; sorting the evaluation results in order of best evaluation result each time prediction and evaluation is performed for one cycle to select a selected number of evaluation result data with best evaluation results; predicting and evaluating a current change for each pattern of the output voltage in the next cycle for each of the selected number of evaluation results; repeating the prediction and evaluation of the current change for each pattern of the output voltage for one cycle and the sorting and selection of the evaluation results until a prediction and evaluation is performed up to a prediction calculation consideration cycle; and outputting data relating to the output voltage of the first H-bridge circuit and the second H-bridge circuit with the best evaluation result for the prediction calculation consideration cycle.
8. The MPC evaluation calculation unit is 8. An active control type magnetic levitation system according to claim 7, wherein when the evaluation results are sorted and selected, the input signals are sorted and selected at once by a sorting network.
9. The MPC evaluation calculation unit is An actively controlled magnetic levitation system as described in claim 7, characterized in that, when sorting and selecting the evaluation results, the sorting network is made multi-staged, the input signals are divided into multiple groups and sorted and selected by multiple sorting networks, and the selected signals are sorted and selected by a sorting network in the next stage.
10. The current change estimation unit 7. An actively controlled magnetic levitation system according to claim 6, characterized in that a current estimation calculation unit that estimates a current one cycle ahead is used to perform a calculation to estimate an output current at time j+1 according to equations (14) and (15), and this calculation is repeated until time j+1 becomes the currently determined output start time, thereby obtaining the output current at the currently determined output start time. ##EQU00014## ##EQU00015## i_j+1: output current at time j+1 i_j: output current at any time j Tc: Output voltage switching period L_load: winding inductance Tdead: Dead time VH_j: the output voltage of the H-bridge circuit corresponding to the state of the output voltage at any time j VH_d: Output voltage during dead time R_load: winding resistance VH_in: positive DC power supply voltage −VH_in: negative DC power supply voltage v_j: state of the output voltage at any time j
11. When the polarity of the first electromagnet winding current is positive and the state value is decreased by one from the previous value, the state is maintained before the transition during the dead time, When the polarity of the first electromagnet winding current is positive and the state value is increased by one from the previous value, the state is set to a state after the transition during the dead time, When the polarity of the first electromagnet winding current is negative and the state value is increased by one from the previous value, the state is maintained before the transition during the dead time, When the polarity of the first electromagnet winding current is negative and the state value is decreased by one from the previous value, an output voltage during the dead time of the first H-bridge circuit is determined as a state after the transition, When the polarity of the second electromagnet winding current is positive and the state value is decreased by one from the previous value, the state before the transition is maintained during the dead time, When the polarity of the second electromagnet winding current is positive and the state value is increased by one from the previous value, the state is set to a state after the transition during the dead time, When the polarity of the second electromagnet winding current is negative and the state value is increased by one from the previous value, the state before the transition is maintained during the dead time, 11. The actively controlled magnetic levitation system according to claim 10, wherein when the polarity of the second electromagnet winding current is negative and the transition is such that the state value is decreased by one from the previous value, the output voltage of the second H-bridge circuit during the dead time is determined as the state after the transition during the dead time.
Citation Information
Patent Citations
Controller of dc / dc converter
JP2018074745A
Controller for DC-DC converter
JP2018074751A
Power conversion system and control method of the same
JP2021170882A
Cited By
Event-triggered robust model prediction control method for suspension system of magnetically levitated train
CN120540085A
