CONTROL DEVICE, DC / DC CONVERSION DEVICE, AND CONTROL METHOD
The DC/DC converter control device uses a low-pass filter and transfer function simulation to align the converter's response with the transfer function, addressing oscillation issues and enabling efficient feedback control constant application and faster transient calculations.
Patent Information
- Application Number
- JP2024504084
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-03-02
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2042-03-02
AI Technical Summary
Existing DC/DC converter control systems struggle to match the response of a transfer function with the actual DC/DC converter, especially when considering disturbances, leading to potential oscillations and mismatched transient responses.
A DC/DC converter control device that includes a low-pass filter to process the output voltage, a control circuit to generate a transfer function simulating the frequency characteristics, and a setting mechanism to adjust the cutoff frequency, ensuring the frequency characteristics of the DC/DC converter and the transfer function align within a specified tolerance.
This approach allows for accurate matching of the DC/DC converter's response with the transfer function, reducing the risk of oscillations and enabling the use of feedback control constants determined from the transfer function in the electrical circuit model, while also allowing for larger time steps in transient calculations.
Smart Images

Figure 0007592215000003 
Figure 0007592215000004 
Figure 0007592215000005
Abstract
Description
[Technical field]
[0001] The present disclosure relates to a control device for a DC / DC converter, a DC / DC conversion device, and a control method for a DC / DC converter. [Background technology]
[0002] The motor control device disclosed in Patent No. 4816754 (Patent Document 1) inputs the transfer function of the controlled object to an inverse model, and filters the output of the inverse model using a first low-pass filter. The cutoff frequency of the first low-pass filter is set to a cutoff frequency that attenuates the observation noise included in the output of the inverse model. [Prior art documents] [Patent documents]
[0003] [Patent Document 1] Patent No. 4816754 Summary of the Invention [Problem to be solved by the invention]
[0004] Incidentally, when evaluating feedback control that takes disturbances into consideration by calculating a transfer function that simulates the frequency characteristics of a DC / DC converter, it is necessary to match the response of the transfer function with the response of the DC / DC converter.
[0005] An object in one aspect of the present disclosure is to provide a technique that can match the response of a DC / DC converter with the response of a transfer function that simulates the DC / DC converter. [Means for solving the problem]
[0006] A control device for a DC / DC converter according to one embodiment includes a low-pass filter that performs low-pass filtering on the output voltage of the DC / DC converter, an operation control unit that controls the operation of the DC / DC converter so as to cause the low-pass filtered output voltage to follow a command value, a generation unit that generates a first transfer function of the DC / DC converter so as to simulate the frequency characteristics of the output voltage of the DC / DC converter, and a setting unit that sets the cut-off frequency of the low-pass filter based on a first frequency range in which the error between the frequency characteristics of the output voltage of the DC / DC converter and the frequency characteristics of the first transfer function is equal to or less than a tolerable value.
[0007] A DC / DC conversion device according to another embodiment includes a DC / DC converter and the above control device.
[0008] A method for controlling a DC / DC converter according to still another embodiment includes the steps of performing low-pass filtering on the output voltage of the DC / DC converter, controlling operation of the DC / DC converter so as to cause the low-pass filtered output voltage to follow a command value, generating a transfer function of the DC / DC converter so as to simulate the frequency characteristics of the output voltage of the DC / DC converter, and setting a cut-off frequency of the low-pass filter that performs the low-pass filtering based on a frequency range in which an error between the frequency characteristics of the output voltage of the DC / DC converter and the frequency characteristics of the transfer function is equal to or smaller than an allowable value. Effect of the Invention
[0009] According to the present disclosure, the response of a DC / DC converter can be matched with the response of a transfer function that simulates the DC / DC converter. [Brief description of the drawings]
[0010] [Figure 1] 1 is a block diagram showing a configuration of a DC / DC conversion device. [Diagram 2] FIG. 2 is a block diagram for explaining a specific configuration of a control circuit. [Diagram 3]FIG. 2 is a block diagram for explaining a method for calculating frequency characteristics of a DC / DC converter. [Figure 4] FIG. 4 is a diagram showing frequency characteristics of an output voltage of a DC / DC converter. [Diagram 5] 11 is a flowchart for explaining a method for generating a transfer function. [Figure 6] 11 is a flowchart illustrating a method for setting the cutoff frequency of a low-pass filter. [Figure 7] FIG. 7 is a diagram showing an example of a result of calculating a cutoff frequency according to the flowchart of FIG. 6. [Figure 8] FIG. 4 is a diagram showing a first example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. [Figure 9] FIG. 11 is a diagram showing a second example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. [Figure 10] FIG. 11 is a diagram showing a third example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. [Figure 11] FIG. 11 is a diagram showing a fourth example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. [Figure 12] FIG. 11 is a diagram showing a fifth example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. [Figure 13] FIG. 11 is a diagram showing a sixth example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. [Figure 14] FIG. 13 is a diagram for explaining a method for setting a cutoff frequency according to a modification of the present embodiment. [Figure 15] FIG. 2 is a configuration diagram of a learning unit related to a DC / DC conversion device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0011] Hereinafter, the present embodiment will be described with reference to the drawings. In the following description, the same components are denoted by the same reference numerals. Their names and functions are also the same. Therefore, detailed description thereof will not be repeated.
[0012] Embodiment 1. <Configuration of the DC / DC Converter> FIG. 1 is a block diagram showing the configuration of the DC / DC converter. Referring to FIG. 1, the DC / DC converter 1 is a power conversion device provided between a DC power supply 2 and a load 3. Specifically, the DC / DC converter 1 includes a DC / DC converter 4 as a main circuit, a control device 8 for the DC / DC converter 4, and a voltage detector 7.
[0013] One end of the DC / DC converter 4 is connected to the DC power supply 2, and the other end is connected to the load 3. The voltage detector 7 detects the output voltage of the DC / DC converter 4 and outputs the detected value of the output voltage to the control device 8 and the frequency analyzer 12. The DC / DC converter 4 may be a bidirectional DC / DC converter.
[0014] The control device 8 includes a control circuit 5, a signal generation circuit 6, and a low-pass filter 10. Each of the control circuit 5, the signal generation circuit 6, and the low-pass filter 10 may be dedicated hardware, or may be a configuration realized by a processor such as a CPU (Central Processing Unit) that executes a program stored in an internal memory (e.g., ROM (Read Only Memory), RAM (Random Access Memory), hard disk, etc.) of the control device 8. For example, when the control circuit 5 is dedicated hardware, the control circuit 5 is composed of an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), or a combination thereof.
[0015] The low-pass filter 10 performs low-pass filter processing on the detected output voltage and outputs the output voltage from which the high-frequency components have been removed to the control circuit 5.
[0016] During normal operation, the control circuit 5 generates a control signal P1 for operating the DC / DC converter 4. Specifically, the control circuit 5 performs various calculations using the output voltage that has been low-pass filtered to generate the control signal P1 for turning on / off the switching element of the DC / DC converter 4. The control circuit 5 outputs the control signal P1 to the DC / DC converter 4.
[0017] The signal generating circuit 6 generates a control signal P2 for operating the DC / DC converter 4 when acquiring the frequency characteristic (i.e., frequency response) of the output voltage of the DC / DC converter 4. The signal generating circuit 6 outputs the control signal P2 to the DC / DC converter 4 based on the sine wave of each frequency received from the frequency analyzer 12 and the control amount.
[0018] The frequency analyzer 12 outputs sine waves of each frequency to the signal generating circuit 6. The frequency analyzer 12 receives an input of the output voltage of the DC / DC converter 4 operated by the control signal P2 generated by the signal generating circuit 6. The frequency analyzer 12 calculates the frequency characteristics of the output voltage and stores it in the database 19.
[0019] 1, the frequency analysis device 12 and the database 19 are configured as external devices to the DC / DC conversion apparatus 1. However, the frequency analysis device 12 and the database 19 may be configured to be included in the DC / DC conversion apparatus 1. The frequency analysis device 12 may be configured to be connected to the DC / DC conversion apparatus 1 when measuring the frequency characteristics. The frequency characteristics of the DC / DC converter 4 may be calculated using an actual device, or may be calculated by simulation using an electric circuit model of the DC / DC converter 4.
[0020] <Control circuit configuration> Fig. 2 is a block diagram for explaining a specific configuration of the control circuit. Referring to Fig. 2, control circuit 5 includes an operation control unit 50, a transfer function generating unit 31, and a setting unit 33. Note that low-pass filter 10 performs low-pass filtering on the output voltage of DC / DC converter 4.
[0021] The operation control unit 50 controls the operation of the DC / DC converter 4 so that the low-pass filtered output voltage follows a command value. Specifically, the operation control unit 50 includes a subtraction unit 13, a feedback control amount calculation unit 14, a limiter processing unit 15, and a control signal generation unit 16.
[0022] The subtraction unit 13 calculates a deviation ΔV (=Vs−V*) by subtracting the low-pass filtered output voltage V* from the command value Vs transmitted from a higher-level device (not shown).
[0023] The feedback control amount calculation unit 14 is composed of a feedback controller (e.g., a P controller, a PI controller, a PD controller, or a PID controller). The feedback control amount calculation unit 14 calculates a control amount by executing feedback control (e.g., P control, PI control, PD control, PID control, etc.) using the deviation ΔV. Typically, the feedback control amount calculation unit 14 generates a control amount X by feedback control for making the deviation ΔV zero.
[0024] The limiter processing unit 15 limits the control amount X to a limit range (i.e., lower limit value: Xd, upper limit value: Xu). Specifically, the limiter processing unit 15 outputs the control amount X set to the upper limit value Xu when the control amount X exceeds the upper limit value Xu, outputs the control amount X set to the lower limit value Xd when the control amount X falls below the lower limit value Xd, and outputs the control amount X when the control amount X is equal to or greater than the lower limit value Xd and equal to or less than the upper limit value Xu. For example, the upper limit value Xu is "1" and the lower limit value Xd is "0".
[0025] The control signal generation unit 16 compares the control quantity X with the carrier signal, and based on the comparison result, generates a control signal P1 as a PWM signal. For example, a triangular wave is used as the carrier signal. The control signal generation unit 16 transmits the control signal P1 to the DC / DC converter 4. The DC / DC converter 4 supplies power from the DC power supply 2 to the load 3 by turning on and off the switching element according to the control signal P1.
[0026] The transfer function generation unit 31 generates a transfer function of the DC / DC converter 4 so as to simulate the frequency characteristics of the output voltage of the DC / DC converter 4 stored in the database 19. Details of the transfer function generation method will be described later.
[0027] The setting unit 33 sets the cut-off frequency Fc of the low-pass filter 10 based on the frequency characteristics of the output voltage of the DC / DC converter 4 and the frequency characteristics of the generated transfer function. The cut-off frequency Fc is output to the low-pass filter 10. Details of the cut-off frequency Fc setting method will be described later.
[0028] <Frequency Characteristics of DC / DC Converter> FIG. 3 is a block diagram for explaining a method for calculating the frequency characteristics of a DC / DC converter. Referring to FIG. 3, the signal generation circuit 6 of the control device 8 includes an addition unit 25 and a control signal generation unit 26.
[0029] The frequency analyzer 12 includes a frequency setting unit 20, a sine wave generation unit 21, and a frequency characteristic calculation unit 22. Each of these functions is realized by a processing circuit. The processing circuit may be dedicated hardware or a processor that executes a program stored in the internal memory of the frequency analyzer 12. When the processing circuit is dedicated hardware, the processing circuit is composed of, for example, an FPGA, an ASIC, or a combination thereof.
[0030] The frequency setting unit 20 calculates a frequency f (e.g., a frequency obtained by equally dividing the frequency range) based on the minimum and maximum frequencies in the frequency range (i.e., the frequency response interval) to be measured and the number of divisions of the frequency range. The frequency setting unit 20 transmits the calculated frequencies f to the sine wave generating unit 21 and the frequency characteristic calculating unit 22 in order from the low frequency side. The frequency setting unit 20 may transmit the current frequency f at the timing when the gain and phase for the previous frequency f are calculated by the frequency characteristic calculating unit 22, or may transmit the frequency f at a fixed time interval set in advance.
[0031] The sine wave generating unit 21 receives a frequency f of an object whose frequency characteristics are to be measured, and outputs a sine wave expressed as "Bsin(2πft)" to the signal generating circuit 6. "B" indicates the amplitude of the sine wave, and "t" indicates time. The amplitude B is set to, for example, about 1 / 10 of the control amount A input to the signal generating circuit 6.
[0032] The adder 25 of the signal generating circuit 6 outputs a value obtained by adding the control amount A and the sine wave generated by the sine wave generating unit 21 to the control signal generating unit 26. The control amount A is a value included in the range from the upper limit value Xu to the lower limit value Xd of the limiter processing unit 15. For example, when the DC / DC converter 4 is a step-down chopper, the control amount A indicates a duty ratio. In this case, the control amount A can range from the upper limit value "1" to the lower limit value "0". When the DC / DC converter 4 is a DAB (Dual Active Bridge) converter, the control amount A indicates a ratio of a phase shift. In this case, the control amount A can range from the minimum value "0" to the maximum value "1".
[0033] The control signal generating unit 26 compares the added value output from the adding unit 25 with the carrier signal, and generates a control signal P2 as a PWM signal based on the comparison result. For example, a triangular wave is used as the carrier signal. The control signal generating unit 16 transmits the control signal P2 to the DC / DC converter 4. The DC / DC converter 4 outputs a voltage by turning a switching element ON / OFF in accordance with the control signal P2.
[0034] The frequency characteristic calculation unit 22 of the frequency analysis device 12 receives the frequency f and the output voltage V of the DC / DC converter 4. When the output voltage V reaches a steady state (for example, when the maximum value of the output voltage V fluctuates within an error of 1%), the frequency characteristic calculation unit 22 calculates the gain and phase corresponding to the component of the received frequency f by performing FFT (Fast Fourier Transformation) analysis on the output waveform. The frequency characteristic calculation unit 22 stores the gain and phase corresponding to the frequency f in the database 19.
[0035] For each frequency f in the frequency range from the lowest frequency to the highest frequency, the frequency analyzer 12 and the signal generating circuit 6 repeatedly execute the above-mentioned process. As a result, the gain and phase corresponding to each frequency f are stored in the database 19, and a file (hereinafter also referred to as "frequency characteristic file") indicating the frequency characteristics (i.e., gain characteristics and phase characteristics) of the output voltage of the DC / DC converter 4 as shown in Fig. 4 is generated. The frequency characteristic file is a file in which the frequency f(i), gain K(i), and phase θ(i) are associated with each number i.
[0036] Fig. 4 is a diagram showing the frequency characteristics of the output voltage of the DC / DC converter. Referring to Fig. 4, a graph 61 shows the gain characteristics, and a graph 62 shows the phase characteristics. Each of the graphs 61 and 62 shown in Fig. 4 is generated based on a frequency characteristics file stored in the database 19.
[0037] In the above, the control device 8 is configured to separately use the signal generating circuit 6 of Fig. 3 in order to obtain the frequency characteristics as shown in Fig. 4, but the present invention is not limited to this configuration. For example, the control signal generating unit 16 of the control circuit 5 may be used to generate the control signal P2.
[0038] 2 are disabled, and the control amount X that has been subjected to limit processing by the limiter processing unit 15 is not input to the control signal generating unit 16. Instead, the adder 25 of the signal generating circuit 6 is provided in the operation control unit 50. The adder 25 inputs an added value (i.e., an added value of the control amount A and the sine wave generated by the sine wave generating unit 21) to the control signal generating unit 16. Then, the control signal generating unit 16 compares the added value with the carrier signal, and generates a control signal P2 based on the comparison result.
[0039] <Transfer function generation method> In this embodiment, a transfer function G1 (for example, the maximum error of the waveform falls within 1%) shown in equation (1) simulating the waveform of the frequency characteristic of DC / DC converter 4 shown in FIG. 4 is considered.
[0040]
number
[0041] Here, “a” in Eq. (1) n , b m , ..., a0, b0" are coefficients. Each coefficient of transfer function G1 is found by applying the least squares method after setting the maximum allowable error so that the maximum error at each frequency between the waveform of the frequency characteristics of DC / DC converter 4 (for example, graphs 61 and 62 in FIG. 4) and the frequency characteristics of transfer function G1 is within 1%.
[0042] Next, a solution (ie, a pole of the transfer function G1) is found that makes the following equation (2) showing the denominator G1b of the transfer function G1 using the coefficients found by the least squares method 0 (ie, G1b=0).
[0043]
number
[0044] If at least one of the signs of the real parts of the poles (i.e., the real parts of the complex solution) is not negative, the value of the transfer function G1 will oscillate in the steady state, and the obtained coefficients cannot be used as the coefficients of the transfer function G1. Therefore, it is necessary to obtain coefficients such that the signs of the real parts of the poles are all negative. The method of generating the transfer function G1 will be specifically described below with reference to the flowchart in FIG. 5.
[0045] Fig. 5 is a flow chart for explaining a method for generating a transfer function. Each process in Fig. 5 is executed by the control circuit 5 (for example, the transfer function generating unit 31).
[0046] The control circuit 5 sets a start number is and an end number ie of frequencies f(i) in the frequency range of the transfer function G1 to be determined (step S100).
[0047] The control circuit 5 sets the gain allowable error εK (%) and the phase allowable error εθ (%) (step S110). The allowable error εK is the allowable error between the gain in the frequency characteristic file of the database 19 and the gain obtained from the transfer function G1. The allowable error εθ is the allowable error between the phase in the frequency characteristic file of the database 19 and the phase obtained from the transfer function G1.
[0048] The control circuit 5 sets the allowable order m0 of the equation (2) of the denominator G1b of the assumed transfer function G1 (step S120). n s n +a n-1 s n-1 The control circuit 5 sets the initial order of the denominator G1b of the transfer function G1 to n=0 (step S130). The control circuit 5 sets the initial order of the denominator G1b of the transfer function G1 to m=1 (step S140). The control circuit 5 reads the frequency f(i), gain K(i), and phase θ(i) from the frequency characteristic file stored in the database 19 (step S150).
[0049] The control circuit 5 judges whether the maximum degree n of the numerator G1a of the transfer function G1 is greater than the maximum degree m of the denominator G1b (i.e., n>m). If n>m (YES in step S160), the control circuit 5 sets the maximum degree of the numerator G1a to n=0 (step S170) and increments the maximum degree of the denominator G1b (i.e., sets m=m+1) (step S190). On the other hand, if n≦m (NO in step S160), the control circuit 5 increments the maximum degree n of the numerator G1a (i.e., sets n=n+1) (step S180).
[0050] Next, the control circuit 5 uses the least squares approximation method to calculate the coefficients of the transfer function G1 of equation (1) to obtain approximate values of the gain K(i) and phase θ(i) of number i (is≦i≦ie) for “ie-is+1” pieces of data from the start number is to the end number ie (step S200).
[0051] The control circuit 5 sequentially calculates the error between the gain K(i) of the frequency characteristic file and the gain of the transfer function G1 using the calculated coefficient for each frequency f(i), and obtains the maximum error Kmax among these errors (step S210).
[0052] The control circuit 5 sequentially calculates the error between the phase θ(i) of the frequency characteristic file and the phase of the transfer function G1 using the calculated coefficient for each frequency f(i), and obtains the maximum error θmax among these errors (step S220).
[0053] The control circuit 5 determines whether or not the condition that the maximum error Kmax of the gain is equal to or less than the allowable error εK (i.e., Kmax≦εK) and the maximum error θmax of the phase is equal to or less than the allowable error εθ (i.e., θmax≦εθ) is satisfied (step S230). If the condition is not satisfied (NO in step S230), the control circuit 5 returns to the process of step S160. If the condition is satisfied (YES in step S230), the control circuit 5 calculates the poles of the transfer function G1 (step S240). Specifically, the control circuit 5 substitutes the calculated coefficients into the formula (2) to calculate the poles that make the denominator G1b=0.
[0054] Next, the control circuit 5 judges whether or not the signs of the real parts of the poles are all negative (step S250). If the signs of the real parts of the poles are all negative (YES in step S250), the control circuit 5 determines the calculated coefficients as the final coefficients of the transfer function G1 (step S270). This allows the transfer function G1 to be determined by simulating the frequency characteristic of the output voltage of the DC / DC converter 4 (i.e., the frequency characteristic file stored in the database 19).
[0055] If not (NO in step S250), the control circuit 5 judges whether the maximum degree m+1 of the denominator G1b is greater than the allowable degree m0 (i.e., m+1>m0) (step S260). If m+1>m0 (YES in step S260), the control circuit 5 ends the process. If m+1≦m0 (NO in step S260), the control circuit 5 returns to the process of step S160.
[0056] As described above, the control circuit 5 (e.g., the transfer function generation unit 31) generates the transfer function G1 so that the gain error is equal to or smaller than the allowable error εK and the phase error is equal to or smaller than the allowable error εθ at each frequency f(i) in the frequency range from the start number is to the end number ie.
[0057] In addition, if the above flowchart ends without determining the coefficients of transfer function G1 (i.e., without executing the processing of step S270), the control circuit 5 again changes the frequency range of transfer function G1 in step S100 (i.e., resets the start number is and the end number ie) and executes the processing from step S110.
[0058] In general, if the frequency of the resonance point is included in the frequency range of the transfer function G1, it becomes difficult to obtain the transfer function G1. Therefore, it is desirable to select a frequency range lower than the frequency at which the resonance occurs.
[0059] <Cutoff frequency setting method> 6 is a flowchart for explaining a method for setting the cutoff frequency of a low-pass filter. Each process in FIG. 6 is executed by the control circuit 5 (for example, the setting unit 33).
[0060] The control circuit 5 sets the end number ie that is the same as the end number ie that was set in step S100 in order to calculate the transfer function G1 in the flowchart of Fig. 5 (step S300). The control circuit 5 sets the gain tolerance εK and the phase tolerance εθ (step S310). The process of step S310 is the same as the process of step S110 in Fig. 5.
[0061] The control circuit 5 sets the coefficients of the transfer function G1 (step S320). Specifically, the coefficients of the transfer function G1 are set to the coefficients calculated in step S270 of Fig. 5. The control circuit 5 reads the frequency f(i), the gain K(i), and the phase θ(i) from the frequency characteristic file stored in the database 19 (step S330).
[0062] The control circuit 5 sets the final number of the data (i.e., frequency f(i), gain K(i), phase θ(i)) of the read frequency characteristic file to ie1 (step S340). The control circuit 5 sets the number i of the data read from the frequency response characteristic file to i=ie (step S350).
[0063] The control circuit 5 determines whether i < ie1 holds (step S360). If i ≥ ie1 (NO in step S360), the control circuit 5 executes the process of step S410 described later. If i < ie1 (YES in step S360), the control circuit 5 increments the number i (i.e., sets i = i + 1) (step S370).
[0064] The control circuit 5 calculates an error Kd between the gain K(i) of the frequency characteristic file and the gain of the transfer function G1 using the coefficient set in step S320 for the frequency f(i) (step S380).
[0065] The control circuit 5 calculates an error θd between the phase θ(i) of the frequency characteristic file and the phase of the transfer function G1 using the coefficient set in step S320 for the frequency f(i) (step S390).
[0066] The control circuit 5 determines whether the condition that the gain error Kd is less than or equal to the allowable error εK (i.e., Kd ≤ εK) and the phase error θd is less than or equal to the allowable error εθ (i.e., θd ≤ εθ) holds (step S400). If the condition holds (YES in step S400), the control circuit 5 returns to the process of step S360. If the condition does not hold (NO in step S400), the control circuit 5 sets the frequency f(i) of number i as the cut-off frequency Fc (step S410) and ends the process.
[0067] FIG. 7 is a diagram showing an example of the result of calculating the cut-off frequency according to the flowchart of FIG. 6. Referring to FIG. 7, a graph 71 showing the gain characteristic of the transfer function G1 is superimposed on a graph 61 of FIG. 4 showing the gain characteristic based on the frequency characteristic file stored in the database 19. Also, a graph 72 showing the phase characteristic of the transfer function G1 is superimposed on a graph 62 of FIG. 4 showing the phase characteristic based on the frequency characteristic file.
[0068] The highest frequency in the frequency range R1 in which the frequency characteristics fall within the allowable error (i.e., Kd≦εK and θd≦εθ) is the frequency f(i) in step S410 of Fig. 6, and this is set as the cutoff frequency Fc. In this way, the cutoff frequency Fc is set to the highest frequency at which the frequency characteristics of the output voltage of the DC / DC converter 4 can be appropriately simulated by the transfer function G1.
[0069] The frequency range R2 from the lowest frequency f(is) to the highest frequency f(ie) is the frequency range used to determine the coefficients of the transfer function G1, as described in the flow chart of Figure 5. It is understood that the frequency range R1 is wider than the frequency range R2 (i.e., the frequency range R2 is included in the frequency range R1).
[0070] <Comparison results> The result of the transient response obtained by simulating an electric circuit simulating the DC / DC conversion device 1 using a circuit simulator is compared with the result of the transient response obtained using the transfer function G1 calculated according to the flowchart of FIG. 5.
[0071] Fig. 8 is a diagram showing a first example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. The vibration waveform confirmed in region 81 in Fig. 8 shows the response waveform of a step input, and the vibration waveform confirmed in region 82 in Fig. 8 is the response waveform when the load is increased in a ramp manner to provide a maximum load, and then the load is reduced in a ramp manner from the maximum load to zero. The load input conditions are the same for Figs. 9 to 11 below.
[0072] Referring to FIG. 8, the dots indicate the transient response of an electric circuit without a low-pass filter, and the solid line indicates the transient response of the transfer function.
[0073] Specifically, the "transient response of the electric circuit" in Fig. 8 is a waveform of the output voltage output from DC / DC converter 4 operating in accordance with control signal P1 when the output voltage that has not passed through low-pass filter 10 in control device 8 in Fig. 2 is input to subtraction unit 13 and the gain of the PI controller constituting feedback control amount calculation unit 14 is 0.01 and the time constant is 0.01.
[0074] The "Transient response of transfer function" in Fig. 8 is a waveform obtained by transient calculation taking into consideration disturbances (load fluctuations) using transfer function Gx (=G1·G2 / (1+G1·G2)) obtained by combining the transfer function of the main circuit (i.e., the DC / DC converter 4) (i.e., the transfer function G1 calculated according to the flowchart in Fig. 5) and the transfer function G2 of the PI controller with the gain set to "0.01" and the time constant set to "0.01".
[0075] It will be understood that under the conditions of FIG. 8, the behavior of the transient response of the electric circuit and the transient response of the transfer function coincide with each other.
[0076] 9 is a diagram showing a second example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. Referring to FIG. 9, the dots indicate the transient response of an electric circuit that does not use a low-pass filter, and the solid line indicates the transient response of the transfer function.
[0077] Specifically, the "transient response of electric circuit" in FIG. 9 is the waveform of the output voltage output from DC / DC converter 4 when the output voltage that has not passed through low-pass filter 10 in control device 8 in FIG. 2 is input to subtraction unit 13, and the gain of the PI controller that constitutes feedback control amount calculation unit 14 is 0.03 and the time constant is 0.01.
[0078] The "Transient response of transfer function" in Figure 9 is a waveform obtained by transient calculation using a transfer function Gx that is a combination of the transfer function G1 of the main circuit and the transfer function G2 of a PI controller with a gain set to "0.03" and a time constant set to "0.01."
[0079] Under the conditions of FIG. 9, it is understood that only the waveform of the electric circuit oscillates, and the transient response of the electric circuit does not match the transient response of the transfer function.
[0080] Fig. 10 is a diagram showing a third example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. Referring to Fig. 10, the dots indicate the transient response of an electric circuit using a low-pass filter, and the solid line indicates the transient response of the transfer function. The cut-off frequency of the low-pass filter is the cut-off frequency Fc calculated by the flowchart of Fig. 6.
[0081] The "transient response of electric circuit" in Figure 10 is the waveform of the output voltage output from DC / DC converter 4 when the output voltage passed through low-pass filter 10 in control device 8 in Figure 2 is input to subtraction unit 13, and the gain of the PI controller constituting feedback control amount calculation unit 14 is 0.03 and the time constant is 0.01.
[0082] The "Transient response of transfer function" in Figure 10 is a waveform obtained by transient calculation using a transfer function Gx that is a combination of the transfer function G1 of the main circuit and the transfer function G2 of a PI controller with a gain set to "0.03" and a time constant set to "0.01."
[0083] It will be understood that under the conditions of FIG. 10, the behavior of the transient response of the electric circuit and the transient response of the transfer function coincide with each other.
[0084] Fig. 11 is a diagram showing a fourth example of a comparison result between the transient response of an electric circuit and the transient response of a transfer function. Referring to Fig. 11, the solid line indicates the transient response of the electric circuit using a low-pass filter, and the circles indicate the transient response of the transfer function.
[0085] Specifically, the "transient response of electric circuit" in FIG. 11 is the waveform of the output voltage output from DC / DC converter 4 when the output voltage passed through low-pass filter 10 in control device 8 in FIG. 2 is input to subtraction unit 13, and the gain of the PI controller constituting feedback control amount calculation unit 14 is 0.06 and the time constant is 0.01.
[0086] The "Transient response of transfer function" in Figure 11 is a waveform obtained by transient calculation using a transfer function Gx that is a combination of the transfer function G1 of the main circuit and the transfer function G2 of a PI controller with a gain set to "0.06" and a time constant set to "0.01."
[0087] Under the conditions of FIG. 11, it is understood that the transient response of the electric circuit and the transient response of the transfer function both oscillate, and their behaviors coincide.
[0088] From the above, when a low-pass filter is not used, the comparison result in Figure 9 shows that the transient response of the electric circuit oscillates, but the transient response of the transfer function does not oscillate, and the behavior of the two does not match. On the other hand, when a low-pass filter is used, the comparison result in Figure 10 shows that the transient response of the electric circuit and the transient response of the transfer function do not oscillate and their behaviors match, and the comparison result in Figure 11 shows that the transient response of the electric circuit and the transient response of the transfer function both oscillate and their behaviors match.
[0089] For this reason, when a low-pass filter is not used, even if a feedback control constant calculated using a transfer function is applied to an electric circuit, there is a high possibility that oscillation will occur in the transient response of the electric circuit. On the other hand, when a low-pass filter is used, if a feedback control constant calculated using a transfer function is applied to the electric circuit, it is presumed that the behavior of the transient response of the electric circuit and the transient response of the transfer function will match.
[0090] FIG. 12 is a diagram showing a fifth example of the comparison result between the transient response of the electric circuit and the transient response of the transfer function. The transient response shown in FIG. 12 is substantially the same as the transient response shown in FIG. 10, but FIG. 12 shows the time step width used in the transient calculation. Specifically, FIG. 12 shows a comparison result in which the time step width used in the transient calculation of the electric circuit is Δt and the time step width used in the transient calculation of the transfer function is 1000 times Δt. Here, Δt=1.0E-06. According to the comparison result, the behavior of the transient response of the electric circuit and the transient response of the transfer function are consistent. From this, it can be understood that the time step width can be made about 1000 times larger in the transient calculation of the transfer function than in the transient calculation of the electric circuit.
[0091] FIG. 13 is a diagram showing a sixth example of the comparison result between the transient response of an electric circuit and the transient response of a transfer function. Referring to FIG. 13, the dots indicate the transient response of an electric circuit using a low-pass filter, and the solid line indicates the transient response of a transfer function. FIG. 13 shows a comparison result when the time step size used in the transient calculation of the transfer function is Δt (however, the same value as Δt in FIG. 12) and the time step size used in the transient calculation of the electric circuit is 10 times Δt (i.e., Δt×10=1.0E-05). According to the comparison result, it is understood that the transient response of the electric circuit causes oscillation. From this, it is understood that the time step size cannot be made too large in the transient calculation of the electric circuit (for example, it cannot be made larger than 1.0E-06).
[0092] From the results in Figure 13, it is estimated that the limit value of the time step size at which the transient response of the electric circuit does not cause oscillation is about "1.0E-06". And, if this time step size is used as the standard, from the results in Figure 12, it is estimated that the time step size in the transient calculation of the transfer function can be about 1000 times larger than in the transient calculation of the electric circuit.
[0093] As described above, by using a transfer function that simulates the DC / DC conversion device 1 that uses the low-pass filter 10 shown in FIG. 1, it is possible to reduce the time required to calculate optimal control constants for feedback control.
[0094] <Modification> In the above-described embodiment, the maximum frequency in the frequency range R1 obtained according to the flowchart in Fig. 6 (i.e., the frequency f(i) in step S410) is set as the cutoff frequency Fc, as shown in Fig. 7. In a modification of this embodiment, another example of the method for setting the cutoff frequency Fc will be described.
[0095] FIG. 14 is a diagram for explaining a method for setting the cutoff frequency according to the modification of the present embodiment.
[0096] In this modification of the present embodiment, the cutoff frequency Fc is set to the maximum frequency of the frequency range R2 used when calculating the transfer function G1 in the flowchart of Fig. 5. This eliminates the need to execute the flowchart of Fig. 6.
[0097] On the other hand, according to the setting method Z2 of the cutoff frequency Fc according to such a modified example, the cutoff frequency Fc is smaller than that according to the setting method Z1 of the cutoff frequency Fc according to Fig. 6. Therefore, when it is desired to simulate the transient response of an electric circuit using a transfer function at a frequency as high as possible, the setting method Z1 of the cutoff frequency Fc according to Fig. 6 should be selected. The person setting the cutoff frequency Fc may arbitrarily select one of the setting methods according to various conditions.
[0098] It is believed that the oscillation conditions caused by the control constant values differ between setting method Z1 and setting method Z2. However, in setting method Z2, as in setting method Z1, the control constants of the feedback control evaluated by the transfer function can be used in the electric circuit model.
[0099] From the above, the control circuit 5 (e.g., setting unit 33) is configured to set the cutoff frequency Fc of the low-pass filter 10 based on the frequency range R1 in which the error between the frequency characteristics of the output voltage of the DC / DC converter 4 and the frequency characteristics of the transfer function G1 is equal to or smaller than the allowable value. Specifically, in one aspect, the control circuit 5 sets the cutoff frequency Fc to the highest frequency in the frequency range R1 using the setting method Z1. In another aspect, the control circuit 5 sets the cutoff frequency Fc to the highest frequency in a frequency range R2 included in the frequency range R1 using the setting method Z2.
[0100] <Advantages> According to the first embodiment, a transfer function that simulates the frequency characteristics of the DC / DC converter 4 is generated, and the cutoff frequency Fc is set based on a frequency range in which the error between the frequency characteristics of the DC / DC converter 4 and the frequency characteristics of the transfer function is equal to or less than a tolerable value. This makes it possible to match the response of the DC / DC converter 4 by feedback control using a low-pass filter in which the cutoff frequency Fc is set with the response of the transfer function. Therefore, the control constants of the feedback control evaluated by the transfer function can be used as they are as control constants in the electric circuit model of the DC / DC converter.
[0101] Furthermore, when performing a transient calculation of an electric circuit including semiconductor circuit elements such as the DC / DC converter 4, it is necessary to set the time step of the transient calculation taking into account the switching frequency of the circuit elements. On the other hand, in the transient calculation of a transfer function, it is not necessary to take into account the switching frequency of the semiconductor circuit elements. Therefore, in the transient calculation of a transfer function, the time step width can be made larger than in the transient calculation of an electric circuit network, and the calculation time can be shortened.
[0102] Embodiment 2 In the second embodiment, a configuration for determining a control constant of a feedback controller using machine learning will be described.
[0103] Fig. 15 is a configuration diagram of a learning unit related to a DC / DC converter. Referring to Fig. 15, learning unit 250 included in control device 8 includes a data acquisition unit 251, a model generation unit 255, a learning result storage unit 260, and an action selection unit 262. Learning unit 250 may be dedicated hardware, or may be configured to be realized by a processor that executes a program stored in an internal memory of control device 8.
[0104] The input parameters of the learning unit 250 are state variables. In this embodiment, the state variables include the maximum error between the output value (i.e., output voltage) of the DC / DC converter 4 shown in FIG. 1 and a target value, and the control constants (e.g., gain and time constant) of the feedback controller constituting the feedback control amount calculation unit 14. The target value is arbitrarily determined by the designer. The output parameters of the learning unit 250 are action values. The action values are control constants that maximize the reward to be obtained in the future.
[0105] When the control constant output from learning unit 250 is set in the feedback controller, control circuit 5 outputs to learning unit 250 the maximum error between the output value and the target value of DC / DC converter 4 during load fluctuation.
[0106] The data acquisition unit 251 acquires learning data including the maximum error between the output voltage of the DC / DC converter 4 and the target voltage, and the control constant of the feedback controller selected by the action selection unit 262 .
[0107] The model generation unit 255 learns an optimal control constant that reduces the maximum error based on learning data including the maximum error and the control constant. That is, the model generation unit 255 uses the learning data to generate a learned model for estimating the optimal control constant from the maximum error.
[0108] The learning algorithm used by the model generation unit 255 may be a known algorithm such as supervised learning, unsupervised learning, or reinforcement learning. As an example, a case where reinforcement learning is applied will be described. In reinforcement learning, an agent (i.e., a subject of action) in a certain environment observes the current state (i.e., environmental parameters) and decides on an action to be taken. The environment changes dynamically due to the agent's actions, and the agent is given a reward according to the change in the environment. The agent repeats this process and learns a course of action that will obtain the most reward through a series of actions.
[0109] Q-learning is known as a representative method of reinforcement learning. In Q-learning, if the action value Q of the action a with the highest Q value at time t+1 is greater than the action value Q of the action a executed at time t, the action value Q is increased, and in the opposite case, the action value Q is decreased. In other words, the action value function Q(s, a) is updated so that the action value Q of the action a at time t approaches the best action value at time t+1. This allows the best action value in a certain environment to be propagated sequentially to the action value in the previous environment.
[0110] As described above, when generating a trained model by reinforcement learning, the model generation unit 255 includes a reward calculation unit 253 and a function update unit 254.
[0111] The reward calculation unit 253 calculates the reward r based on a reward criterion that specifies how to increase or decrease the reward used in learning the learning model, a control constant, and a maximum error. The reward criterion is set so that the reward is increased when the maximum error is reduced, and the reward is reduced when the maximum error is increased. Specifically, when the acquired control constant is set in a state indicated by the acquired maximum error, the reward calculation unit 253 increases the reward (for example, gives a reward of "+1") if the maximum error is reduced, and reduces the reward (for example, gives a reward of "-1") if the maximum error is increased.
[0112] The function update unit 254 updates a function for determining an output parameter (i.e., a control constant) of the learning unit 250 according to the reward calculated by the reward calculation unit 253, and outputs the updated function to the learning result storage unit 260. For example, in the case of Q-learning, the action value function Q(s t ,a t ) is used as a function to calculate the control constant. t represents the state of the environment at time t, and a t represents the action at time t.
[0113] The action selection unit 262 selects a control constant based on the action value function Q(s, a), which is the learning result of the learning unit 250 stored in the learning result storage unit 260. Typically, the action selection unit 262 randomly selects a control constant in the initial state. In the middle of the learning stage, the action selection unit 262 selects a control constant using a known action selection method (e.g., the ε-greedy method). The control constant selected by the action selection unit 262 becomes the output parameter of the learning unit 250.
[0114] The control circuit 5 calculates the maximum error when the control constant output from the learning unit 250 (i.e., the control constant selected by the action selection unit 262) is set in the feedback controller, and outputs the maximum error to the learning unit 250. The data acquisition unit 251 acquires the maximum error and the control constant selected by the action selection unit 262, and outputs them to the model generation unit 255.
[0115] The learning unit 250 repeats the above learning. When the action-value table is no longer updated, the function update unit 254 updates the action-value function Q(s t ,a t When the converged action-value function Q(s) has converged (for example, when the maximum error is within 0.1%), the learning unit 250 determines that learning has ended. In this case, the action selection unit 262 selects the converged action-value function Q(s) as the t ,a t) is used to select the control constant that will provide the greatest reward (i.e., the optimal control constant is output). Furthermore, the learning result storage unit 260 stores a learned model for estimating the optimal control constant from the maximum error.
[0116] As described above, in order to generate a trained model for determining optimal control constants, it is necessary to repeatedly perform training, which takes a significant amount of time, and therefore it is preferable to shorten the training time as much as possible.
[0117] As described in the first embodiment, the time step size can be made larger in the transient calculation of the transfer function than in the transient calculation of the electric circuit. Therefore, it is considered that the learning time can be reduced by outputting the maximum error calculated using the transfer function to the learning unit 250.
[0118] In this case, the state variables include the maximum error between the output value and the target value of the transfer function Gx obtained by combining the transfer function G1 of the main circuit (i.e., the DC / DC converter 4) and the transfer function G2 of the feedback controller constituting the feedback control amount calculation unit 14, and the control constant of the feedback controller. Note that the output parameter of the learning unit 250 as the action value is the control constant.
[0119] When the control constant output from the learning unit 250 is set in the feedback controller, the control circuit 5 outputs the maximum error between the output value of the transfer function Gx and the target value when a disturbance corresponding to a load change is applied to the learning unit 250. The data acquisition unit 251 of the learning unit 250 acquires learning data including the maximum error between the output value of the transfer function Gx and the target value and the control constant of the feedback controller. The subsequent process flow of the learning unit 250 is the same as that described above.
[0120] Since the time step size of the transient calculation of the transfer function is about 1000 times that of the transient calculation of the electric circuit, it is expected that the learning time will be shortened to about 1 / 1000. For example, if the learning time when using the electric circuit is one day (i.e., 1440 minutes), the learning time when using the transfer function will be 1.4 minutes, which is a significant reduction.
[0121] <Advantages> According to the second embodiment, by learning the optimal control constant of the feedback controller using a transfer function, the learning time for generating a learned model for calculating the control constant can be significantly reduced.
[0122] Other embodiments. The configurations exemplified as the above-mentioned embodiments are examples of the configurations of the present disclosure, and may be combined with other known technologies, or may be modified, such as by omitting some parts, without departing from the scope of the present disclosure. In addition, the above-mentioned embodiments may be implemented by appropriately adopting the processes and configurations described in other embodiments.
[0123] The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present disclosure is defined by the claims, not the above description, and is intended to include all modifications within the scope and meaning equivalent to the claims. [Explanation of symbols]
[0124] 1 DC / DC conversion device, 2 DC power supply, 3 load, 4 converter, 5 control circuit, 6 signal generation circuit, 7 voltage detector, 8 control device, 10 low-pass filter, 12 frequency analysis device, 13 subtraction unit, 14 feedback control amount calculation unit, 15 limiter processing unit, 16, 26 control signal generation unit, 19 database, 20 frequency setting unit, 21 sine wave generation unit, 22 frequency characteristic calculation unit, 25 addition unit, 31 transfer function generation unit, 33 setting unit, 50 operation control unit, 250 learning unit, 251 data acquisition unit, 253 reward calculation unit, 254 function update unit, 255 model generation unit, 260 learning result storage unit, 262 behavior selection unit.
Claims
1. A control device for a DC / DC converter, a low-pass filter that performs low-pass filtering on an output voltage of the DC / DC converter; an operation control unit that controls an operation of the DC / DC converter so that the output voltage processed by the low-pass filter follows a command value; a generator that generates a first transfer function of the DC / DC converter so as to simulate a frequency characteristic of an output voltage of the DC / DC converter; a setting unit that sets a cutoff frequency of the low-pass filter based on a first frequency range in which an error between a frequency characteristic of an output voltage of the DC / DC converter and a frequency characteristic of the first transfer function is equal to or smaller than a tolerable value.
2. The control device according to claim 1 , wherein the generation unit generates the first transfer function such that the error is equal to or smaller than the allowable value at each frequency in a second frequency range included in the first frequency range.
3. The control device according to claim 1 , wherein the setting unit sets the cutoff frequency to a maximum frequency in the first frequency range.
4. The control device according to claim 2 , wherein the setting unit sets the cutoff frequency to a highest frequency in the second frequency range.
5. the generation unit generates a composite transfer function by combining the first transfer function and a second transfer function of a feedback controller included in the operation control unit; a data acquisition unit that acquires learning data including a maximum error between an output value and a target value of the composite transfer function and a control constant of the feedback controller; The control device according to any one of claims 1 to 4, further comprising: a model generation unit that generates a learned model for estimating the control constant from the maximum error using the learning data.
6. A DC / DC converter; A DC / DC conversion device comprising the control device according to any one of claims 1 to 5.
7. A method for controlling a DC / DC converter, comprising: performing low pass filtering on the output voltage of the DC / DC converter; controlling an operation of the DC / DC converter so as to make the low-pass filtered output voltage follow a command value; generating a transfer function of the DC / DC converter so as to simulate a frequency characteristic of an output voltage of the DC / DC converter; and setting a cutoff frequency of a low-pass filter that executes the low-pass filter processing based on a frequency range in which an error between the frequency characteristics of the output voltage of the DC / DC converter and the frequency characteristics of the transfer function is equal to or less than an allowable value.
Citation Information
Patent Citations
JP1973016754B1
Power supply
JP2000324810A
Controller for DC-DC converter
JP2007068253A
Transfer function generation based on pulse width modulation information
JP2016516383A
JPP7010415B
