Power conversion device

Abstract

The main circuit (2) includes a switching element, and the main circuit (2) converts input power and supplies the power to a load (5). The detection unit (6) detects an output value of the main circuit (2). The output value of the main circuit (2) is an output voltage of the main circuit (2) or an output current of the main circuit (2). The control device (3) controls the main circuit (2). The control device (3) switches a control mode of the main circuit (2) from a first control mode to a second control mode at a first time point at which the output value starts to change, and switches the control mode of the main circuit (2) from the second control mode to the first control mode at a second time point at which a change in a direction of the output value is judged based on a detection value obtained by the detection unit (6).

Description

TECHNICAL FIELD

[0001] The present application relates to a power conversion device. BACKGROUND

[0002] The feedback control device described in Patent Literature 1 has a first control section that outputs an operation amount for a control object with information related to a control deviation between a measured value of the control object and a target value as input, a second control section that includes a learning control section that determines a parameter for outputting the operation amount for the control object by machine learning with information related to the control deviation as input, an addition operation section that adds a first operation amount output from the first control section and a second operation amount output from the second control section and outputs the added operation amount to the control object, and a limiter that limits the second operation amount output from the second control section.

[0003] Patent Literature 1: Japanese Patent Application Publication No. 2019-71405 SUMMARY

[0004] For example, in a case where the control object is an output voltage or an output current, the output voltage or the output current is varied due to various reasons such as load variation or noise. The magnitude of the variation and the operation amount required for control differ depending on the magnitude of the load variation or the like that is generated. In the feedback control device described in Patent Literature 1, since the second operation amount output from the second control section is limited by the limiter, it is sometimes difficult to reduce the variation of the output voltage or the output current in a case where the output voltage or the output current is greatly varied.

[0005] Therefore, an object of the present application is to provide a power conversion device that can reduce the variation of an output voltage or an output current.

[0006] The power conversion device of the present application has a main circuit that includes a switching element, the main circuit converting input power to supply to a load, a detection section that detects an output value of the main circuit, and a control device that controls the main circuit. The output value of the main circuit is an output voltage of the main circuit or an output current of the main circuit. The control device switches the control mode of the main circuit from a first control mode to a second control mode at a first time point at which the output value starts to vary, and switches the control mode of the main circuit from the second control mode to the first control mode at a second time point at which a switching of the direction of variation of the output value is determined based on a detection value obtained by the detection section.

[0007] EFFECT OF THE INVENTION

[0008] According to the present application, the control device switches the control mode of the main circuit from the first control mode to the second control mode at a first time point at which the output value starts to change, and switches the control mode of the main circuit from the second control mode to the first control mode at a second time point at which it is determined based on the detection value obtained by the detection section that the direction of the change in the output value is switched. Thereby, the variation in the output voltage or the output current can be reduced. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 is a diagram showing the structure of the power conversion device 1 of Embodiment 1.

[0010] Figure 2 is a diagram showing a step-down chopper as one example of the main circuit 2.

[0011] Figure 3 is a flowchart showing the processing flow of the estimation section 12.

[0012] Figure 4 is a diagram showing a neural network that predicts the output voltage.

[0013] Figure 5 is a diagram showing an example of the variation in the output voltage Vo.

[0014] Figure 6 is a diagram showing the circuit model in which the circuit of Figure 2 is detailed in order to accurately calculate the conduction rate D of the step-down chopper.

[0015] Figure 7 is a diagram showing the simulated waveforms.

[0016] Figure 8 is a diagram showing five control methods.

[0017] Figure 9 is a diagram showing the waveforms of the output voltage Vo and the conduction rate D when multiple learning is performed.

[0018] Figure 10 is a diagram showing the structure of the power conversion device 21 of Embodiment 2.

[0019] Figure 11 is a diagram showing the structure of the control device 3 in the case where the functions of the control device 3 are implemented using software. DETAILED DESCRIPTION

[0020] Hereinafter, the embodiments will be described with reference to the drawings.

[0021] Embodiment 1

[0022] Figure 1 is a diagram showing the structure of the power conversion device 1 of Embodiment 1.

[0023] Power conversion device 1 includes a main circuit 2, a control device 3, and an output voltage detection unit 6. Main circuit 2 converts power input from power supply 4 and supplies the converted power to load 5. Control device 3 controls main circuit 2. Output voltage detection unit 6 detects the output voltage of main circuit 2.

[0024] Power conversion device 1 is used, for example, in a headlamp or laser processing equipment. When power conversion device 1 is used in a headlamp, load 5 is a light-emitting element such as an LED. When power conversion device 1 is used in a laser processing equipment, load 5 is a laser diode or a discharge load.

[0025] Main circuit 2 is a switching circuit. The switching circuit includes a switching element driven by a PWM (Pulse Width Modulation) signal generated by control device 3. The switching element can be a FET (Field Effect Transistor) or an IGBT (Insulated Gate Bipolar Transistor). As long as main circuit 2 is a switching circuit, any structure is acceptable.

[0026] The main circuit 2 may also be a step-up chopper or a step-up / step-down chopper. It may also be an isolated circuit such as a flyback converter, an LLC converter, or a DAB (Dual Active Bridge) converter. It may also be an AC / DC converter circuit that converts AC to DC, or a DC / AC converter circuit that converts DC to AC.

[0027] Figure 2 1 is a diagram showing a step-down chopper as an example of the main circuit 2 .

[0028] The step-down chopper includes a switching element Ta, diodes D1 and D2, capacitors Ca and Cf, and a coil Lf. Capacitor Ca is placed between input node ND1 and input node ND2. Capacitor Cf is placed between output node ND3 and output node ND4. Diode D2 is placed between intermediate node ND5 and intermediate node ND6. Diode D1 and switching element Ta are placed in parallel between input node ND1 and intermediate node ND5. Coil Lf is placed between intermediate node ND5 and output node ND3.

[0029] The output voltage detection unit 6 detects the output voltage Vo of the main circuit 2 and outputs the detected output voltage Vo to the control device 3 .

[0030] The control device 3 includes an ADC 11 , an estimating unit 12 , a main circuit control unit 13 , and a PWM generating unit 14 .

[0031] The ADC 11 generates an output voltage detection value Vdt by converting the output voltage detected by the output voltage detection section 6 into a digital value at a sampling cycle Ts. The ADC 11 outputs the output voltage detection value Vdt to the estimation section 12 and the main circuit control section 13.

[0032] The estimation section 12 judges whether or not a variation (hereinafter, referred to as an output voltage variation) of the output voltage Vo of the main circuit 2 occurs, based on the output voltage command value Vr and the output voltage detection value Vdt output by the ADC 11. The estimation section 12 outputs the correction information Cr to the main circuit control section 13 in a case where it is judged that the output voltage variation occurs. Further, the output voltage command value Vr can be input from the outside or can be stored internally by the control device 3.

[0033] The main circuit control section 13 performs normal feedback control, band correction feedback control, and integral term adjustment of feedback control, based on the output voltage command value Vr, the output voltage detection value Vdt, and the correction information Cr.

[0034] The PWM generation section 14 generates a PWM signal for driving the main circuit 2, based on the output of the main circuit control section 13.

[0035] The control device 3 generates a PWM signal for driving the main circuit 2 in such a manner that the output voltage Vo of the main circuit 2 follows the output voltage command value Vr. The control device 3 increases the responsiveness when the output voltage variation occurs, as compared with the normal time. The high responsiveness when the output voltage variation occurs is realized by the estimation section 12 and the main circuit control section 13.

[0036] Next, details of the estimation section 12 will be described.

[0037] The estimation section 12 outputs the output voltage prediction value Vep from the time point at which the variation of the output voltage Vo is detected to the time point at which the direction of the variation of the output voltage Vo is switched, and outputs "0" as the correction information Cr in other cases.

[0038] The estimation section 12 generates the output voltage prediction value Vep based on the past and current output voltage detection values Vdt. The estimation section 12 predicts the switching of the direction of the variation of the output voltage Vo using the output voltage prediction value Vep. The direction of the variation of the output voltage Vo refers to a direction in which the output voltage Vo rises (positive direction) or a direction in which the output voltage Vo falls (negative direction). The variation direction switching refers to the switching of the variation direction from the positive direction to the negative direction or from the negative direction to the positive direction. The estimation section 12 can accurately predict the occurrence of the switching of the direction of the variation of the output voltage Vo by using the output voltage prediction value Vep, and thus can prevent a large correction from being continuously applied for a long time in the band correction feedback control of the main circuit control section 13 described later.

[0039] Figure 3 is a flowchart showing the processing flow of the estimation section 12.

[0040] In step S101, the estimation section 12 acquires the output voltage detection value Vdt and the output voltage command value Vr.

[0041] In step S102, the estimation section 12 judges whether or not the output voltage fluctuation has occurred, based on the output voltage detection value Vdt. The judgment of the output voltage fluctuation can be any method using the output voltage detection value Vdt. For example, it can be that the estimation section 12 judges that the output voltage fluctuation has occurred when the output voltage detection value Vdt continuously changes in a direction in which the difference from the output voltage command value Vr is enlarged a plurality of times. Or, it can be that the estimation section 12 judges that the output voltage fluctuation has occurred when the difference between the output voltage detection value Vdt and the output voltage command value Vr is larger than or equal to a threshold value. The estimation section 12 changes the flag flg to "1" if it judges that the output voltage fluctuation has occurred. Further, in S102, the estimation section 12 does not change the flag flg to 0. The estimation section 12 changes the flag flg to 0 in S105 described later. In the case where the flag flg is already "1", step S102 can be omitted.

[0042] In step S103, the estimation section 12 discriminates whether the flag flg is "0" or "1". In the case where the flag flg is "0", the processing proceeds to step S107, and in the case where the flag flg is "1", the processing proceeds to step S104.

[0043] In S104, the estimation section 12 predicts the output voltage Vo after X sampling periods and (X+1) sampling periods from the current. The estimation section 12 can predict the output voltage Vo using an equation obtained by modeling the power conversion device 1, or using a function in which the output voltage waveform at the time of the output voltage fluctuation has been simulated in advance, or using a machine learning model trained. In the following description, a neural network is used as the machine learning model, but it is not limited thereto, and can be a support vector machine, a random forest, or a gradient boosting, etc.

[0044] Figure 4 is a diagram showing a neural network that predicts the output voltage.

[0045] The neural network has 3 layers in common. The number of units is 3, 6, 2 in the order of the input layer, the intermediate layer (hidden layer), and the output layer. The number of layers of the intermediate layer of the neural network can also be larger than or equal to 2 layers, and the number of units of each layer can be arbitrary, but since the control of the power conversion device 1 requires high-speed processing, a small-scale model as shown in FIG. 8 is preferable. Figure 4

[0046] ​Next, the operation of the neural network will be described.

[0047] The output voltage detection values N[n-2], N[n-1], N[n] of three sampling points, the current and the past two points, are input to the input layer of the neural network. n denotes the current sampling time point. From the output layer, the output voltage prediction value N[n+X] after X sampling periods and the output voltage prediction value N[n+X+1] after (X+1) sampling periods are output, which are estimated from the output voltage detection values of the three sampling points input to the input layer.

[0048] For example, X=1 can be set, but X=1 is not limited. The value of X is preferably set taking into account system delay such as delay due to feedback control and detection. For example, if the system has a delay of three sampling periods, X=3 can be set. Alternatively, instead of inputting the current output voltage detection value to the input layer of the neural network, the output voltage detection values of the past three points can be input.

[0049] In step S105, the estimation section 12 uses the output voltage prediction values N[n+X] and N[n+X+1] calculated in step S104 to determine whether or not the direction of change of the output voltage Vo has changed.

[0050] Figure 5 is a graph showing an example of change in the output voltage Vo. Here, X=1 is described.

[0051] In Figure 5 , the output voltage Vo initially changes in the direction of decreasing. Therefore, if the output voltage prediction value N[n+1] after one sampling period and the output voltage prediction value N[n+2] after two sampling periods are compared, N[n+1]>N[n+2].

[0052] In Figure 5 , if time elapses, the direction of change switches to the direction of increasing the output voltage Vo. That is, if the output voltage prediction value N[n+1] after one sampling period and the output voltage prediction value N[n+2] after two sampling periods are compared, N[n+1]<N[n+2]. In this case, the estimation section 12 determines that the direction of change of the output voltage Vo has changed.

[0053] The estimation section 12 changes the flag flg to "0" at the timing at which it is determined that the direction of the variation in the output voltage Vo is switched. In this way, the reason why the control is switched by the control device 3 at the timing at which it is determined that the direction of the variation in the output voltage Vo is switched, instead of at the timing at which the direction of the variation in the output voltage Vo is switched, is that there is a delay in the PWM generation section 14 for generating the PWM signal after the feedback operation is performed by the main circuit control section 13, a response delay in the main circuit 2, and the like. The control device 3 switches the control in advance in anticipation of these delays.

[0054] In step S106, when the flag flg is "0", the process proceeds to step S107, and when the flag flg is "1", the process proceeds to step S108.

[0055] In step S107, the estimation section 12 outputs "0".

[0056] In step S108, the estimation section 12 outputs the output voltage prediction value N[n+X] after X sampling periods, which is predicted in S104, as the correction information Cr. In addition, the value output by the estimation section 12 in S108 can be, for example, the output voltage prediction value N[n+X+1] after (X+1) sampling periods, as long as it is a value predicted in step S104.

[0057] The order of the steps of the flowchart shown can also be exchanged. Figure 3 For example, step S104 can be executed before step S102, but from the viewpoint of the amount of calculation, it is preferable to execute step S104 after step S102 as shown in Figure 3 In addition, in step S104, any number of points can be predicted as long as two consecutive points are included. In this case, the switching determination in step S105 can be arbitrary as long as it uses the output voltage prediction value calculated in step S104. That is, it can also be deformed in accordance with the number of output voltage prediction values.

[0058] In addition, the determination of whether or not the output voltage variation occurs can also not be performed in S102, but can be performed based on a signal transmitted from the outside (not shown). For example, if it is a system in which the output voltage command value Vr is periodically changed, the output voltage variation occurs in accordance with the change in the command value, so a signal corresponding to the period of the change can also be prepared in advance, and the output voltage variation can be determined based on the signal.

[0059] In step S104, the output voltage prediction value N[n+X] is calculated using Figure 4In the case of the neural network, the output voltage detection value before 1 sampling period and the output voltage detection value before 2 sampling periods need to be saved. Therefore, for example, in step S101, the estimation unit 12 can also save the output voltage detection value before 1 sampling period and the output voltage detection value before 2 sampling periods.

[0060] In the determination of the generation of the switching of the variation direction in step S105, the output voltage detection value (actual sampling value) can also be used. For example, a determination achieved based on a comparison of the output voltage detection value at the current sampling and the output voltage detection value before 1 sampling, or a determination achieved based on a comparison of the output voltage detection value and the output voltage prediction value, and the like can be used in combination with the determination achieved based on a comparison of the two output voltage prediction values explained in step S105. In addition, the output voltage prediction value can be only the point of the output voltage prediction value Vep[n+X] after X sampling periods, and the generation of the switching of the variation direction can be determined by comparing the output voltage prediction value after X sampling periods estimated before 1 sampling period and the output voltage prediction value after X sampling periods estimated in the current sampling period.

[0061] Next, details of the main circuit control unit 13 will be described.

[0062] The main circuit control unit 13 controls the main circuit 2 in the first control mode at normal times. The first control mode is, for example, normal feedback control. The main circuit control unit 13 controls the main circuit 2 in the second control mode from the time when the output voltage varies until the time when it is determined based on the output voltage prediction value that the switching of the variation direction of the output voltage has occurred. The second control mode is, for example, a feedback control with correction. The main circuit control unit 13 controls the main circuit 2 in the first control mode from the time when it is determined based on the output voltage prediction value that the switching of the variation direction of the output voltage has occurred, and adjusts the integral term of the feedback control.

[0063] The responsiveness of the power conversion device 1 at the time of the variation of the output voltage can be improved by switching the control mode at the time of the variation of the output voltage. By adjusting the integral term at the time of recovery after the switching of the control mode, it is possible to prevent the operation after the switching from becoming unstable.

[0064] First, the basic operation of the main circuit control unit 13, that is, the normal feedback control will be described.

[0065] The main circuit control section 13 calculates a difference value SA of the output voltage detection value Vdt and the output voltage command value Vr. The main circuit control section 13 determines a control amount CON output to the PWM generation section 14 in such a manner that the difference value SA becomes zero. The control based on the control amount CON can be a classical control such as PI (Proportional Integral) control or PID (Proportional Integral Derivative) control, or a modern control such as H∞ control, or a control using machine learning such as fuzzy control. In the following description, an example using PID control is described.

[0066] Next, switching of the control mode at the time of output voltage variation is described.

[0067] The main circuit control section 13 judges whether or not the output voltage variation has occurred based on the output of the estimation section 12. The main circuit control section 13 judges that no output voltage variation has occurred in the case where the output of the estimation section 12 is "0", and controls the main circuit 2 in the first control mode (normal feedback control). Equation (Al) shows an example of a calculation formula of the normal feedback control.

[0068] [Equation 1]

[0069] N on = N B-K p (N[n]-N R )-K I ∑(N[n]-N R )-K D (N[n]-N[n-1]) …(Al)

[0070] The main circuit control section 13 judges that the output voltage variation has occurred in the case other than where the output of the estimation section 12 is "0", and switches the control mode to the second control mode. Here, as an example of the second control mode, a corrected feedback control using the output voltage predicted value N[n+X] after X sampling periods sent from the estimation section 12 as the correction information Cr is described. Equation (1) shows an example of a calculation formula of the corrected feedback control. The proportional term PT is represented by equation (la). The integral term IT is represented by equation (lb). The differential term DT is represented by equation (lc).

[0071] [Equation 2]

[0072] N on = B -K p {N[n]-(N R +ΔN)}-K I∑(N[n]-N R )-K D (N[n]-N[n-1])...(1)

[0073] PI=K p {N[n]-(N R +ΔN)}...(1a)

[0074] IT=K I ∑(N[n]-N R )...(1b)

[0075] DI=K D (N[n]-N[n-1])...(1c)

[0076] n represents the current sampling period. N[n] represents the current output voltage detection value. N[n-1] represents the output voltage detection value before 1 sample. N on is the output of the main circuit control section 13, i.e., the control amount, K p is the proportional gain, K I is the integral gain, and K D is the differential gain. N B and N R are the bias term and the reference value, respectively. N B and N R represent the output voltage command value Vr. ΔN is the correction term using the output voltage prediction value N[n+X] after X samples, and is represented by equation (2).

[0077] [Mathematical Expression 3]

[0078] ΔN=N R -N[n+X]...(2)

[0079] In the correction feedback control shown in equation (1) and equation (2), the output voltage prediction value N[n+X] is corrected with respect to the output voltage command value Vr represented by the reference value N R and the bias term N B . For example, when the value of the output voltage Vo decreases due to output voltage fluctuation, an action equivalent to increasing the output voltage command value Vr is performed. Alternatively, when the value of the output voltage Vo increases due to output voltage fluctuation, an action equivalent to decreasing the output voltage command value Vr is performed. As described above, the correction feedback control can improve the responsiveness of the feedback control when the output voltage fluctuation occurs by correcting the output voltage command value Vr when the output voltage fluctuation occurs.

[0080] By setting the value of equation (2) to "0" when the output of the estimation section 12 is "0", the normal feedback control can be implemented without changing the mathematical expression.

[0081] Equations (1), (2) are shown as examples of the band correction feedback control, but equations other than them can be used, for example, an equation without the bias term N B and the reference value N R using the output voltage command value Vr.

[0082] The point at which the output voltage prediction value is used to perform correction in the band correction feedback control can be a point other than the output voltage command value Vr, for example, the output voltage prediction value can be used to correct the feedback control gain such as the proportional gain Kp.

[0083] In addition, instead of using the output voltage prediction value, control can be switched to a predetermined feedback control gain during a period in which generation of the output voltage variation is detected. However, using the output voltage prediction value has the advantage that the amount of correction can be adjusted in correspondence with the size of the generated output voltage variation.

[0084] Next, adjustment of the integral term IT of the feedback control is described.

[0085] In the present embodiment, at the timing at which the output voltage variation direction switching is judged by the estimation section 12, the main circuit control section 13 switches the control mode of the main circuit 2 from the second control mode (band correction feedback control) to the first control mode (normal feedback control), and adjusts the integral term IT of the feedback control.

[0086] At the timing at which the output voltage variation direction switches, the slope of the output voltage Vo is approximately 0. When the slope of the output voltage Vo is 0, a state in which the power that the main circuit 2 attempts to supply to the load 5 and the power consumed by the load 5 are balanced is reached, and a state that is the same as the steady state is reached. Therefore, the main circuit control section 13 calculates the conduction rate D based on an equation that represents the relationship between the input voltage and the output voltage and the conduction rate D in the steady state at the timing at which the output voltage variation direction switching is judged. The main circuit control section 13 adjusts the integral term IT of the feedback control so that the conduction rate D becomes the calculated value.

[0087] An example in the case of the step-down chopper in which the main circuit 2 is Figure 2 described.

[0088] In the step-down chopper, when the slope of the output voltage Vo is 0, no current flows through the capacitor Cf, and the size of the current flowing through the coil Lf is equal to the size of the current flowing through the load 5. This is because the current flowing through the coil Lf when the switching element Ta is on and the current flowing through the coil Lf when the switching element Ta is off are balanced, and therefore the conduction rate D can be calculated by Equation (3).

[0089] [Math. 4]

[0090]

[0091]

[0092] Vin represents the input voltage of the main circuit 2, and L represents the reactance value of the coil Lf. The conduction rate D is calculated by the main circuit control section 13 from the input voltage value and the output voltage value (the output voltage detection value or the output voltage prediction value) stored in advance by the equation (3). The main circuit control section 13 adjusts the integral term IT of the feedback control so that the PWM signal generated by the PWM generation section 14 becomes the signal of the calculated conduction rate D.

[0093] In the case where the main circuit 2 is a step-up chopper, the conduction rate D can be calculated by the equation (3a). In the case where the main circuit 2 is a step-up / down chopper, the conduction rate D can be calculated by the equation (3b). The main circuit control section 13 adjusts the integral term IT of the feedback control based on the calculated conduction rate D at the timing of the switching of the variation direction of the output voltage. By adjusting the integral term IT of the feedback control, it is possible to prevent the control from becoming unstable after the switching from the band correction feedback control to the normal feedback control.

[0094] [Math. 5]

[0095]

[0096]

[0097] As explained in the above example, in the case where the output voltage prediction value is used for the correction of the output voltage command value Vr to perform the band correction feedback control, it is equivalent to inputting the output voltage command value Vr different from the actual one. As a result, after the end of the band correction feedback control, an unintended value is sometimes set to the integral term IT, and thus the adjustment of the integral term IT is effective.

[0098] Further, it is also possible to adjust the integral term IT a plurality of times with respect to one output voltage variation. For example, the estimation section 12 detects the switching of the variation direction of the output voltage three times from the time point at which the output voltage variation occurs, and outputs the integral term adjustment command signal to the main circuit control section 13 each time the switching of the variation direction of the output voltage is detected. The main circuit control section 13 adjusts the integral term IT each time the integral term adjustment command signal is received.

[0099] The value of the input voltage Vin required for calculating the conduction rate in the voltage reduction chopper can be stored in advance in the control device 3, or can be detected by a detection unit not shown. The conduction rate D can be calculated based on a circuit model for calculating the conduction rate D.

[0100] Figure 6 is a circuit model for detailing the circuit of Figure 2 .

[0101] Figure 6 The circuit model of Figure 2 is a model in which the resistance component r1 of the switching element Ta, the resistance component r2 of the diode D2, and the resistance component r L of the coil Lf are added to the circuit of

[0102] [mathematical expression 6]

[0103]

[0104]

[0105] The output current Io in the expression (4) can be detected by an output current detection circuit not shown, or can be estimated from the output voltage Vo that has been detected.

[0106] The cause of the variation in the output voltage includes variation in the resistance value of the load 5, variation caused by noise, variation caused by a change in the output command value, and the like. Here, as one example, a method of estimating the output current Io from the output voltage Vo in the case where the output voltage varies due to variation in the resistance value of the load 5 will be described.

[0107] First, the main circuit control unit 13 calculates the output current value Iost before the output voltage varies, based on the expression (5) obtained by modifying the expression (4).

[0108] [mathematical expression 7]

[0109]

[0110] Dst, Vinst, and Vost are the conduction rate of the main circuit 2, the input voltage of the main circuit 2, and the output voltage of the main circuit 2, respectively, before the output voltage varies. Vinst can be detected by a detection circuit not shown, or can be stored in advance in the internal of the main circuit control unit 13. Dst can be calculated from the control amount actually output by the main circuit control unit 13.

[0111] Next, the main circuit control section 13 calculates the current Ic flowing through the capacitor Cf at the time of the generation of the output voltage variation. The current Ic is calculated by using the output voltage detection values Vo[m-1], Vo[m] detected by the sampling before and after the generation of the output voltage variation, the sampling period Ts, and the capacitance C of the capacitor Cf by the equation (6).

[0112] [Equation 8]

[0113]

[0114] As shown in the equation (7), the main circuit control section 13 calculates the sum of the output current value Iost before the generation of the output voltage variation and the current Ic flowing through the capacitor Cf at the time of the generation of the output voltage variation, thereby calculating the output current Ioc after the generation of the output voltage variation.

[0115] [Equation 9]

[0116] I oc = O ost + I c (7)

[0117] As shown in the equation (8), the main circuit control section 13 calculates the resistance value Rc of the load 5 after the generation of the output voltage variation based on the output current Ioc after the generation of the output voltage variation and the output voltage detection value Vo[m].

[0118] [Equation 10]

[0119]

[0120] The main circuit control section 13 estimates the value of the output current Io using the resistance value Rc of the load after the generation of the output voltage variation and the output voltage detection value Vdt. The output current estimation is used for the calculation of the on-duty D of the switching element 3. Figure 6 The equation for calculating the on-duty D of the switching element 3 is shown in the equation (9).

[0121] [Equation 11]

[0122]

[0123] As described above, even if the output current Io is not detected, the on-duty D can be calculated by estimating the output current Io. By estimating the output current Io, the on-duty D can be calculated with high accuracy without adding a circuit for detecting the output current.

[0124] In the present embodiment, the output voltage prediction value Vep calculated by the estimation unit 12 can also be used for the main circuit control unit 13. By using the same output voltage prediction value Vep for the output voltage variation direction switching determination in the estimation unit 12 and the band correction feedback control in the main circuit control unit 13, the amount of calculation of the control device 3 can be reduced, and the power conversion device can be downsized and simplified.

[0125] Next, the operation and effects of the present embodiment are described using circuit simulation by Myway Plus Co., Ltd.

[0126] The simulation model of the power conversion device 1 of Figure 1 was created. Figure 1 The main circuit 2 of the power conversion device 1 of Figure 2 uses a step-down chopper as shown in FIG. 6. The input voltage Vin of the main circuit 2 was set to 20 [V], and the output voltage command value Vr was set to 10 [V]. PID control was used for feedback control. The neural network shown in FIG. 7 was used for prediction of the output voltage variation. In the band correction feedback control, the output voltage command value Vr was corrected using the output voltage prediction value estimated by the neural network. The load 5 was a resistive load, and in order to reproduce the output voltage variation, the resistance value of the load 5 was changed from 50 [Ω] to 20 [Ω] halfway through the simulation. Figure 4

[0127] Figure 7 is a graph showing the simulation waveforms.

[0128] In Figure 7 (a) and (b), the horizontal axis is time. In Figure 7 (a), the vertical axis is the output voltage Vo. In Figure 7 (b), the vertical axis is the duty ratio D. In Figure 7 (a), the waveforms W1 to W5 of the output voltage Vo in the five control methods Cm1 to Cm5 are shown. In Figure 7 (b), the waveforms P1 to P5 of the duty ratio D in the five control methods Cm1 to Cm5 are shown.

[0129] Figure 8 is a graph showing the five control methods.

[0130] In the control method Cm1, the main circuit control unit 13 only performs the usual feedback control, and does not perform the band correction feedback control, the integral term adjustment of the feedback control, and the determination of the switching of the variation direction of the output voltage.

[0131] ​The control method Cm2 is a control method of the embodiment. In the control method Cm2, the main circuit control section 13 performs the correction feedback control in addition to the usual feedback control, and performs the integral term adjustment of the feedback control. Further, the main circuit control section 13 judges the switching of the variation direction of the output voltage by comparing the output voltage predicted value after one sampling period and the output voltage predicted value after two sampling periods which are estimated by the neural network.

[0132] In the control method Cm3, the main circuit control section 13 performs the correction feedback control in addition to the usual feedback control, but does not perform the integral term adjustment of the feedback control. Further, the main circuit control section 13 judges whether or not the switching of the variation direction of the output voltage occurs by comparing the output voltage detection values at two sampling time points.

[0133] In the control method Cm4, the main circuit control section 13 performs the correction feedback control in addition to the usual feedback control, and performs the integral term adjustment of the feedback control. Further, the main circuit control section 13 judges whether or not the switching of the variation direction of the output voltage occurs by comparing the output voltage detection values at two sampling time points.

[0134] In the control method Cm5, the main circuit control section 13 performs the correction feedback control in addition to the usual feedback control, and performs the integral term adjustment of the feedback control. The main circuit control section 13 judges whether or not the switching of the variation direction of the output voltage occurs by comparing the output voltage detection value at one sampling time point and the output voltage predicted value after one sampling period which is estimated by the neural network.

[0135] In the A timing, the resistance value of the load 5 is changed from 50 [Ω] to 20 [Ω]. Thus, the output voltage variation in which the output voltage Vo is decreased occurs.

[0136] In the B timing, in the control methods Cm2, Cm3, Cm4, Cm5, it is judged that the output voltage variation occurs, and the control mode is switched from the usual feedback control to the correction feedback control. It is confirmed that the duty D is changed to a large value by the correction feedback control.

[0137] In the C timing, in the control method Cm2 in which the switching of the variation direction is judged by the comparison of the two output voltage prediction values, the band correction feedback control is ended, the integral term IT of the feedback control is adjusted, and the normal feedback control is executed. If the waveform W2 of the output voltage Vo of the control method Cm2 is confirmed, the switching timing of the variation direction is approximately coincident with the C timing. That is, it can be confirmed that the switching judgment of the variation direction can be accurately performed by the comparison of the two output voltage prediction values. Further, it can be confirmed that the conduction rate D after the switching to the normal feedback control by the adjustment of the integral term of the feedback control does not become an abnormal value, but is adjusted to a value close to the final value.

[0138] In the D timing, in the control methods Cm3, Cm4, Cm5, the switching of the variation point is judged, and the band correction feedback control is ended. The timing at which the switching of the variation point is judged is later than the switching timing of the variation point of the waveforms W3 to W5 of the output voltage Vo of the actual control methods Cm3 to Cm5. It is known that in the control methods Cm3, Cm4, Cm5, a large conduction rate D is given for a long time compared with the control method Cm2 in which the switching of the variation point is accurately judged.

[0139] Further, if the waveforms W1 to W5, P1 to P5 of the control methods Cm1 to Cm5 are compared, the following matters can be observed.

[0140] In the waveforms W2 to W5 of the control methods Cm2 to Cm5, the variation of the output voltage Vo in the lower limit direction is reduced compared with the waveform W1 of the control method Cm1, and the reason is that a large conduction rate D is given by the band correction feedback control.

[0141] In the control methods Cm3 to Cm5 in which the occurrence of the switching of the variation direction is judged using the output voltage detection value, a large conduction rate D is given for a longer time than the switching of the variation direction compared with the control method Cm2 in which the occurrence of the switching of the variation direction is judged by the comparison of the two output voltage prediction values. As a result, it can be confirmed that a large kick voltage is generated in the upper limit direction of the output voltage Vo.

[0142] In the control methods Cm4, Cm5, the integral term IT of the feedback control is adjusted compared with the control method Cm3. As a result, it can be confirmed that the maximum value of the output voltage Vo can be suppressed in the waveforms W4, W5 of the output voltage Vo. The difference between the control methods Cm4 and Cm5 is the difference between the judgment of the switching of the variation point by the comparison of the two output voltage detection values and the judgment of the switching of the variation point by the comparison of the output voltage detection value and the output voltage prediction value obtained by the neural network.

[0143] Thus, the control method of the present embodiment implements the integral term adjustment of the feedback control and the correction feedback control, and judges the switching timing of the variation direction of the output voltage by comparing the two output voltage prediction values. The control method of the present embodiment can reduce the decrease of the output voltage, and also prevents the overshoot voltage. Thus, the control method of the present embodiment is a control method capable of making the output voltage follow the output voltage command value at the highest speed.

[0144] Next, the learning of the neural network in the case where the neural network is used in S104 of the estimation section 12 will be described.

[0145] The neural network is previously learned using the waveform data at the time of the output voltage variation, and adjusts the weighting coefficients between the units. The values of the current and past two sampling points of the waveform at the time of the output voltage variation (N[n], N[n-1], N[n-2]) are input, and the values after X sampling periods and (X+1) sampling periods (N[n+X], N[n+X+1]) are used as teaching data. The weighting coefficients are adjusted in a manner that enables high-precision estimation of the values after X sampling periods and (X+1) sampling periods (N[n+X], N[n+X+1]) from the values of the current and past two sampling points (N[n], N[n-1], N[n-2]).

[0146] As for the range of the output voltage waveform at the time of the load variation used for the learning of the neural network, in order to make the judgment of the generation of the variation direction switching with high precision, it is preferable to set a period longer than the period from the generation of the variation to the switching timing of the load variation.

[0147] Further, although it is described that the weighting coefficients are learned in advance, it can also be configured to learn simultaneously with the estimation operation of the neural network, and update the weighting coefficients successively.

[0148] The learning of the neural network can also be performed by an external device. The estimation section 12 acquires the waveform of the output voltage detection value at the time of the output voltage variation, and outputs it to the external device. The external device learns the neural network based on the received waveform, and outputs the adjusted weighting coefficients to the estimation section 12. The estimation section 12 sets the adjusted weighting coefficients to the neural network, and performs the estimation operation.

[0149] In the present embodiment, the main circuit 2 is controlled by the correction feedback control that corrects the output voltage command value using the output voltage prediction value output by the neural network. Therefore, the waveform of the output voltage prediction value of the neural network is different from the actual output voltage waveform. Therefore, by repeating the cycle of the acquisition of the output voltage waveform, the learning of the neural network, the prediction of the output voltage by the neural network, the correction feedback control, the acquisition of the output voltage waveform, the learning of the neural network such a plurality of times, it is possible to obtain the weighting coefficient of the neural network that can greatly reduce the variation of the output voltage.

[0150] Figure 9 is a graph showing the waveforms of the output voltage Vo and the conduction rate D at the time when the learning is performed a plurality of times.

[0151] In Figure 9 In (a), (b), the horizontal axis is time. In Figure 9 In (a), the vertical axis is the output voltage Vo. In Figure 9 In (b), the vertical axis is the conduction rate D.

[0152] If referring to Figure 9 (a), (b), it is possible to suppress the variation of the output voltage Vo when the learning is performed (Y+1) times compared to when the learning is performed Y times (Y represents an integer). Also, it is confirmed that the variation of the output voltage Vo can be suppressed when the learning is performed (Y+2 times) compared to when the learning is performed (Y+1) times.

[0153] Further, in the case where the learning is performed a plurality of times, it is also preferable to perform the judgment of the generation of the switching of the direction of the variation of the output voltage realized by the comparison of the output voltage prediction value after X sampling periods and the output voltage prediction value after (X+1) sampling periods and the integral term adjustment. Thereby, it is possible to prevent the operation of the power conversion device 1 from becoming unstable after the correction feedback control ends.

[0154] Embodiment 2

[0155] Figure 10 is a graph showing the structure of the power conversion device 21 of Embodiment 2.

[0156] Figure 10 The power conversion device 21 illustrated in Figure 1 is different from the power conversion device 1 of Figure 10 The power conversion device 21 of Figure 10 is different from the power conversion device 1 of Figure 1 In Figure 1The same structure and the structure element performing the same action.

[0157] The power conversion device 1 of Embodiment 1 controls the output voltage Vo in accordance with the output voltage command value Vr and suppresses the output voltage variation. The power conversion device 21 of Embodiment 2 controls the output current Io in accordance with the output current command value Ir and suppresses the output current variation.

[0158] The control of the output voltage Vo and the control of the output current Io are different as follows. In the output voltage control, if the load 5 is changed from 50 [Ω] to 20 [Ω], the output voltage Vo generates a variation in the direction of becoming small. In the output current control, if the load 5 is changed from 50 [Ω] to 20 [Ω], the output current Io generates a variation in the direction of becoming large. The actions at the time of the same load variation are opposite. The reason is that, in the voltage control, the side of the resistance value of the load which is small becomes the heavy load, but in the current control, the side of the resistance value of the load which is large becomes the heavy load.

[0159] As described above, the power conversion devices of Embodiments 1 and 2 control the main circuit by the band correction feedback control in the period from the detection of the generation of the output voltage variation or the output current variation to the switching of the variation direction of the output voltage based on the detected value of the output voltage or the output current of the main circuit. By the band correction feedback control, the responsiveness of the power conversion device can be improved, and thus the variation amplitude of the output voltage variation or the output current variation can be reduced. Thereby, the high performance of the power conversion device can be achieved, and the miniaturization and the low cost of the output filter of the main circuit can be achieved.

[0160] The power conversion devices of Embodiments 1 and 2 judge the switching of the variation direction of the output voltage variation or the output current variation using the output voltage predicted value or the output current predicted value after X sampling periods and (X+1) sampling periods. By using the predicted value, the generation of the switching of the variation direction can be accurately judged in a state where the delay factors such as the detection delay and the control delay do not affect. By accurately judging, the period of the band correction feedback control can be appropriately set. Thereby, the hunting of the output voltage or the output current due to the band correction feedback control performed in a period longer than the necessary period can be prevented, and thus the stability of the power conversion device can be improved.

[0161] The power conversion device of Embodiment 1 and Embodiment 2 performs the band correction feedback control using the output voltage prediction value or the output current prediction value, and corrects the output voltage command value or the output current command value. By using the output voltage prediction value or the output current prediction value, a correction value that is appropriate to the magnitude of the output voltage variation or the output current variation that occurs can be given, and thus the responsiveness improvement and the stability improvement of the power conversion device can be achieved.

[0162] The power conversion device of Embodiment 1 and Embodiment 2 adjusts the integral term of the feedback control at the timing of switching from the band correction feedback control to the ordinary feedback control. Since the integral term is adjusted at the timing of switching the variation direction of the output voltage variation or the output current variation, a simple conduction ratio formula in the steady state of the main circuit can be used. As a result, the amount of calculation of the control device can be reduced, and thus the size reduction and the cost reduction of the control device can be achieved. Further, by adjusting the integral term of the feedback control at the timing of switching from the band correction feedback control to the ordinary feedback control, the variation of the output voltage or the output current that occurs when a value largely deviating from the ideal value is set to the integral term after the switching can be prevented, and thus the stability of the power conversion device can be improved.

[0163] The power conversion device of Embodiment 1 and Embodiment 2 performs both the band correction feedback control and the judgment of the switching of the variation direction of the output voltage variation or the output current variation using the output voltage prediction value or the output current prediction value after X sampling periods and after (X+1) sampling periods. Thus, the amount of operation of the control device can be reduced, and thus the size reduction and the cost reduction of the control device can be achieved.

[0164] The control device 3 described in Embodiments 1 and 2 can also perform the equivalent operation by hardware or software of a digital circuit.

[0165] Figure 11 is a diagram showing the configuration of the control device 3 in the case where the functions of the control device 3 are implemented using software. The control device 3 has a processor 5002 and a memory 5001 connected to a bus 5003. The processor 5002 executes a program stored in the memory 5001.

[0166] It should be understood that the embodiments disclosed herein are illustrative and not restrictive in all aspects. The scope of the present application is not represented by the above description but by the claims, and is intended to include all modifications equivalent in meaning and scope to the scope of the claims.

[0167] Explanation of Reference Numerals

[0168] 1, 21 power conversion device, 2 main circuit, 3 control device, 4 power supply, 5 load, 6 output voltage detection part, 12 estimation part, 13 main circuit control part, 14 PWM generation part, 22 output current detection part, 5001 memory, 5002 processor, 5003 bus, D1, D2 diode, Lf coil, Ta switching element.

Claims

1. A power conversion device comprising: a main circuit including a switching element, the main circuit converting input power and supplying the converted power to a load; a detection unit that detects an output value of the main circuit; and a control device for controlling the main circuit, The output value of the main circuit is the output voltage of the main circuit or the output current of the main circuit, The control device switches the control mode of the main circuit from the first control mode to the second control mode at a first time point when the output value begins to change, predicts the output value of the main circuit at a time after the detection time point of the detection value based on the detection value obtained by the detection unit, and switches the control mode of the main circuit from the second control mode to the first control mode at a second time point when it is determined based on the predicted value of the output value of the main circuit that a change direction of the output value has changed. The first control method is feedback control for controlling the output value so as to follow the command value. The second control method is feedback control with correction in which correction is added to the feedback control based on a predicted value of the output value of the main circuit.

2. The power conversion device according to claim 1, wherein: Based on the detection value, the control device predicts the output value of the main circuit in two timed sampling cycles after the detection time point of the detection value, and based on the predicted value of the output value of the main circuit in the two sampling cycles, determines that a switch in the change direction of the output value has occurred.

3. The power conversion device according to claim 1, wherein: The command value of the feedback control with correction is a command value corrected based on a predicted value of the output value of the main circuit.

4. The power conversion device according to claim 1, wherein: The control gain of the feedback control with correction is a control gain corrected based on a predicted value of the output value of the main circuit.

5. The power conversion device according to any one of claims 1, 3 and 4, wherein: The control device calculates the conduction ratio of the main circuit at the second time point, and adjusts the integral term of the feedback control based on the conduction ratio.

6. The power conversion device according to claim 5, wherein: The control device calculates the conduction ratio of the main circuit at the second time point according to an equation representing a relationship between an input voltage and an output voltage in a steady state and the conduction ratio.

7. The power conversion device according to any one of claims 1 to 4, wherein: The control device detects a change in the output value based on a detection value obtained by the detection unit.

8. The power conversion device according to claim 2, wherein: The control device predicts the output value of the next sampling cycle and the output value of the next two sampling cycles based on the detection value of the current sampling cycle, the detection value of the previous sampling cycle, and the detection values ​​of the previous two sampling cycles.

9. The power conversion device according to any one of claims 1, 3 and 4, wherein: The feedback control is PI control or PID control.

10. The power conversion device according to claim 1 or 2, wherein: The control device uses the trained machine learning model to predict the output value of the main circuit at a timing after the detection time point of the detection value based on the detection value.

11. The power conversion device according to claim 10, wherein: The machine learning model is a neural network.

Images