Power conversion device and estimation device

Abstract

A main circuit (50A) is configured to include a switching element (51), a reactor (53), and a capacitor (54), and converts power from a power source (2) into supply power to a load (5). A control unit (150) controls the on / off duty ratio (DTY) of the switching element (51) so as to control an output from the main circuit (50A) to the load (5). An estimation unit (200) is configured to: acquire data containing a detected value (Iref) of the reactor current (IL), a detected value (Voref) of the output voltage (Vo), and the duty ratio (DTY) of the switching element (51) periodically for a plurality of cycles; obtain estimated values of state coefficients in a state equation of the main circuit (50A) by regression calculation using the acquired data; and further calculate an estimated value of the circuit parameter of each component in the main circuit (50A) using the estimated values of the state coefficients.

Description

Power conversion device and estimation device

[0001] This disclosure relates to a power conversion device and a device for estimating the circuit parameters thereof.

[0002] Methods have been disclosed for estimating the parameters of components mounted on a power converter and for optimizing the control of the power converter or detecting abnormalities in the power converter according to the estimated parameters. For example, Japanese Patent Application Publication No. 2021-121143 (Patent Document 1) describes a control method that estimates LC filter constants based on measurements taken at a certain point in the operation of a DC-DC converter circuit, and sets the switching frequency to optimize the ripple voltage based on the estimated capacitance and inductance.

[0003] Furthermore, Japanese Patent Publication No. 2013-031235 (Patent Document 2) describes a control method in which a control function equation defining the relationship between the output voltage and the derivative of the output voltage of a switching power supply is defined in advance, and the input voltage and output voltage are sampled at a timing synchronized with the switching period of the main switching element to calculate the on time and off time of the main switching element so as to satisfy the said control function equation. Patent Document 2 describes that in this control method, the parameters of the circuit constants are periodically estimated and the circuit constants are periodically updated.

[0004] Japanese Patent Publication No. 2021-121143 Japanese Patent Publication No. 2013-031235

[0005] In Patent Document 1, when estimating the inductance and capacitance values ​​of the inductor and capacitor constituting an LC filter, in addition to the input voltage and output voltage, the ripple voltage of the output voltage and the ripple current flowing through the inductor are required as data. However, ripple voltage and ripple current cannot be detected by sampling at simple timings of voltage and current; accurate detection values ​​(sampling values) at the timings when the voltage and current are at their maximum and minimum are required. Therefore, there are concerns that acquiring input for parameter estimation will become costly due to the need for high-speed detectors and computational processing for accurate timing prediction.

[0006] Furthermore, while Patent Document 2 performs parameter estimation of circuit constants from the output voltage sampling value, the parameter estimation targets a predefined composite coefficient that includes the product of the inductance value and the capacitance value. Therefore, in order to utilize the parameter estimation results for anomaly detection or output control, it is desirable to estimate the parameters of individual components, but Patent Document 2 does not allow for the independent estimation of inductance and capacitance values. Consequently, there is a concern that the effectiveness of parameter estimation will be limited.

[0007] This disclosure was made to solve these problems, and its purpose is to estimate the circuit parameters of each component of a power converter using easily obtainable detection values.

[0008] In certain aspects of this disclosure, a power converter is provided. The power converter converts power from a power source into power supplied to a load. The power converter comprises a main circuit connected between the power source and the load, a current detector, a first voltage detector, a control unit, and an estimation unit. The main circuit comprises one or more reactors or transformers, one or more switching elements, and one or more capacitors. The current detector detects the current flowing through the reactor or transformer. The first voltage detector detects at least one of the output voltage from the main circuit to the load or the voltage of the capacitor. The control unit controls the on / off state of the switching elements to control the output from the main circuit to the load. The estimation unit estimates the circuit parameters of the main circuit. During a data acquisition period including the non-steady-state operation of the power converter, the estimation unit acquires data including current detection values ​​from the current detector, voltage detection values ​​from the first voltage detector, and time ratios of the switching elements for multiple periods corresponding to the switching period of the switching elements or the calculation period of the control unit. Furthermore, the estimation unit is configured to obtain estimated values ​​for each state coefficient of the state equation of the main circuit by regression calculation using the acquired data, and then calculate estimated values ​​for the circuit parameters using these estimated state coefficients. The circuit parameters include at least one of the following, which are included in any of the state coefficients: the inductance value of a reactor or transformer, the resistance value of a reactor or transformer, the resistance value of a switching element, the capacitance value of a capacitor, and the resistance value of a capacitor.

[0009] In another aspect of this disclosure, a device for estimating the circuit parameters of a power converter is provided. The power converter converts power from a power source into power supplied to a load. The power converter comprises a main circuit connected between the power source and the load, a current detector, a first voltage detector, and a control unit. The main circuit comprises one or more reactors or transformers, one or more switching elements, and one or more capacitors. The current detector detects the current flowing through the reactor or transformer. The first voltage detector detects at least one of the output voltage from the main circuit to the load or the voltage across the capacitors. The control unit controls the on / off state of the switching elements to control the output from the main circuit to the load. The estimation device is configured to perform the following processes during the data acquisition period, including the non-steady-state operation of the power converter: acquiring data including current detection values ​​from a current detector, voltage detection values ​​from a first voltage detector, and time ratios of switching elements for multiple periods at intervals corresponding to the switching period of the switching elements or the calculation period of the control unit; obtaining estimated values ​​of each state coefficient of the state equation of the main circuit by regression calculation using the acquired data; and further calculating estimated values ​​of circuit parameters using the estimated values ​​of the state coefficients. The circuit parameters include at least one of the following that are included in any of the state coefficients: inductance value of a reactor or transformer, resistance value of a reactor or transformer, resistance value of a switching element, capacitance value of a capacitor, and resistance value of a capacitor.

[0010] According to this disclosure, estimated values ​​of circuit parameters can be calculated by obtaining estimated values ​​of each state coefficient in the state equation of the main circuit through regression calculations using current detection values, voltage detection values, and the time ratio of switching elements for multiple periods. Therefore, the circuit parameters of each component of the power converter can be estimated using easily obtainable detection values.

[0011] This is a block diagram showing the configuration of the power converter according to this embodiment. This is a circuit block diagram illustrating the configuration of the power converter according to Embodiment 1. This is a conceptual waveform diagram illustrating PWM control. This is a conceptual diagram illustrating an example of a state equation. This is a circuit diagram illustrating the analysis model of the main circuit (buck chopper) shown in Figure 2. This is the equivalent circuit diagram of Figure 5 during the ON period of the switching element. This is the equivalent circuit diagram of Figure 5 during the OFF period of the switching element. This is a circuit diagram illustrating the analysis model of the main circuit according to the first configuration example of Embodiment 2. This is a circuit diagram illustrating the analysis model of the main circuit according to the second configuration example of Embodiment 2. This is a circuit diagram illustrating the analysis model of the main circuit according to the third configuration example of Embodiment 2. This is the equivalent circuit diagram of the main circuit shown in Figure 9. This is a circuit diagram illustrating the analysis model of the main circuit according to the fourth configuration example of Embodiment 2. This is a circuit diagram illustrating the analysis model of the main circuit according to the fifth configuration example of Embodiment 2. This is an example of an operating waveform diagram of the main circuit shown in Figure 12. This is a circuit diagram illustrating an example configuration of the average value acquisition circuit according to Embodiment 3. This is a waveform diagram for illustrating the sampling timing of the ADC according to Embodiment 3. This is a conceptual diagram illustrating an example configuration of the neural network according to Embodiment 3. This is a circuit block diagram illustrating a first configuration example of the power conversion device according to Embodiment 4. This is a circuit block diagram illustrating a second configuration example of the power conversion device according to Embodiment 4. This is a block diagram illustrating a first arrangement example of the estimation device according to Embodiment 5. This is a block diagram illustrating a second arrangement example of the estimation device according to Embodiment 5.

[0012] Embodiments of this disclosure will be described in detail below with reference to the drawings. In the following, the same or corresponding parts in the drawings will be denoted by the same reference numerals, and their descriptions will not be repeated in principle.

[0013] Embodiment 1. Figure 1 is a block diagram showing the configuration of the power conversion device 10 according to this embodiment.

[0014] As shown in Figure 1, the power converter 10 according to this embodiment comprises a main circuit 50 connected to a power source 2 and supplying power to a load 5, and a computing device 100. The power converter 10 converts power from the power source 2 into power supplied to the load 5.

[0015] The main circuit 50 is a switching circuit that includes one or more coils or transformers, one or more capacitors, and one or more switching elements, as will be shown in numerous examples later. The switching elements are driven by control signals from the control unit 150 for PWM (Pulse Width Modulation) control, etc. (hereinafter also simply referred to as "PWM signals"). For example, the switching elements can be made of any element that can be switched on and off in response to the control signal, such as an IGBT (Insulated Gate Bipolar Transistor) or an FET (Field Effect Transistor) with diodes connected in antiparallel.

[0016] As will be described later, the main circuit 50 is equipped with one or more current detectors for detecting the current flowing through a coil or transformer, and one or more voltage detectors for detecting one or more voltages.

[0017] The configuration of the main circuit 50 is arbitrary, as long as it is a power conversion circuit (switching circuit) that includes a switching element. As will be described later, the main circuit 50 may be composed of non-isolated power converters such as step-down choppers, step-up choppers, and step-up / step-down choppers, or it may be composed of isolated power converters such as flyback converters, LLC resonant converters, and DAB (Double Active Bridge) converters. Alternatively, the main circuit 50 may be composed of a multi-phase circuit such as an interleaved circuit, or a multi-level circuit, an AC / DC converter that converts AC to DC, or a DC / AC converter that converts DC to AC.

[0018] The arithmetic unit 100 may be configured such that at least some of its functions are realized by predetermined arithmetic processing using digital electronic circuits such as FPGAs (Field Programmable Gate Arrays), and / or by hardware processing such as the operation of analog electronic circuits such as comparators, operational amplifiers, and differential amplifier circuits. Alternatively, the arithmetic unit 100 may be configured such that at least some of its functions are realized by predetermined arithmetic processing using software processing by executing a pre-stored program.

[0019] The arithmetic unit 100 includes a control unit 150 for controlling the main circuit 50 and an estimation unit 200 for estimating the parameters of the components constituting the main circuit 50. The respective functions of the control unit 150 and the estimation unit 200, which will be described later, can be realized by hardware processing and / or software processing by the arithmetic unit 100.

[0020] The control unit 150 generates a PWM signal to supply the desired power to the load 5 using at least one detected value detected by the main circuit 50. The main circuit 50 can perform power conversion between the power supply 2 and the load 5, along with output control to the load 5, by controlling the on / off state of the switching elements in accordance with the PWM signal.

[0021] The estimation unit 200 uses at least one current detection value, a voltage detection value, and the time ratio of the PWM signal detected by the main circuit 50 to estimate circuit parameters, which are constant values ​​of the components constituting the main circuit 50 (inductance value, capacitance value, resistance value, etc.). The circuit parameter information estimated by the estimation unit 200 can be transmitted to the control unit 150 and / or the power converter 10 either as is or after signal processing.

[0022] (Circuit Configuration Example) The following describes an example of the configuration of the power conversion device according to this embodiment.

[0023] Figure 2 is a circuit block diagram illustrating an example configuration of Embodiment 1 of the power converter 10. Referring to Figure 2, the power converter 10 according to Embodiment 1 comprises a main circuit 50A connected between a power supply 2 composed of a DC power supply and a load 5 composed of a resistor, and a calculation device 100. The calculation device 100 includes a control unit 150 and an estimation unit 200.

[0024] The main circuit 50A is composed of a step-down chopper including a switching element 51, a diode 52, a reactor 53, and a capacitor 54, and can supply an output voltage Vo (Vo ≤ Vi) obtained by stepping down the input voltage Vi from the power supply 2 to the load 5. Furthermore, the main circuit 50 is equipped with a current detector 71 that detects the reactor current IL flowing through the reactor 53 and a voltage detector 72 that detects the output voltage Vo.

[0025] The control unit 150 includes an ADC (Analog to Digital Converter) 152, a control calculation unit 154, and a PWM generation unit 156. The ADC 152 converts the reactor current IL (analog value) detected by the current detector 71 and the output voltage Vo (analog value) detected by the voltage detector 72 into digital values ​​at a predetermined period and outputs them as the output voltage detection value Voref and the current detection value Iref. This period can be determined to correspond to the switching period, which is the reciprocal of the switching frequency of the switching element 51, or to the calculation period of the control unit 150.

[0026] The control calculation unit 154 receives an output voltage detection value Voref from the ADC 152 and calculates the time ratio DTY of the switching element 51, which is a control variable, so that the voltage deviation (Vout* - Voref) between the output voltage detection value Voref and an internally held target value (for example, the output voltage target value Vout*) becomes zero. For example, the control calculation unit 154 can calculate the time ratio DTY by a PI (proportional-integral) control calculation that takes the above voltage deviation as input.

[0027] Furthermore, the calculation of the controlled variable (time ratio DTY) by the control calculation unit 154 is not limited to the above example, as long as it can supply the desired power to the load 5 according to the target value. For example, the control calculation unit 154 may perform the control calculation using two values: an output voltage detection value Voref and a current detection value Iref, and the target value may be input from outside the calculation device 100. In addition, the controlled variable (time ratio DTY) may be calculated using other classical control methods such as PID (proportional-differential-integral) control, or modern control methods such as H∞ (H-infinity) control, or model predictive control, rather than being limited to PI control.

[0028] The PWM generation unit 156 generates a PWM signal Spwm for on / off control of the switching element 51 by PWM control based on the time ratio DTY calculated by the control calculation unit 154. The time ratio DTY is updated with each calculation cycle.

[0029] Figure 3 shows a conceptual waveform diagram illustrating PWM control for generating a PWM signal. As shown in Figure 3, the time ratio DTY is calculated by the control calculation unit 154 within the range of 0 ≤ DTY ≤ 1.0. The PWM generation unit 156 generates a PWM signal Spwm based on the comparison result between the carrier wave CW, whose value periodically changes between 0 and 1.0, and the time ratio DTY from the control calculation unit 154. The carrier wave CW can be composed of a triangular wave or a sawtooth wave, as exemplified in Figure 3.

[0030] During periods when DTY ≥ CW, the PWM signal Spwm is set to a high level, and during periods when DTY < CW, the PWM signal Spwm is set to a low level.

[0031] Referring again to Figure 2, the switching element 51 is turned on during the high-level period of the PWM signal Spwm and turned off during the low-level period. Therefore, the period Ts of the carrier wave CW corresponds to the switching period Ts of the switching element 51. Also, as shown in Figure 3, it is understood that the time ratio DTY corresponds to the time ratio of the ON period Yon to the switching period Ts of the switching element 51.

[0032] The estimation unit 200 includes a data holding unit 210 and an estimation calculation unit 220. The data holding unit 210 receives the output voltage detection value Voref and the current detection value Iref from the ADC 152, and the time ratio DTY from the control calculation unit 154, thereby acquiring and storing information necessary for circuit parameter estimation (data for regression calculation, described later). The estimation calculation unit 220 holds a circuit parameter estimation formula for the main circuit 50 that has been prepared in advance, and calculates the circuit parameter estimation value PCest using the result of a regression calculation using the data stored in the data holding unit 210 and the said circuit parameter estimation formula. In the example in Figure 2, the circuit parameter estimation value PCest may include estimated values ​​such as the inductance value L of the reactor 53 and the capacitance value C of the capacitor 54 of the main circuit 50A (step-down chopper).

[0033] The estimated circuit parameter value PCest can be output to the outside of the arithmetic unit 100 (power converter 10) either as is or after signal processing. Alternatively, the estimated circuit parameter value PCest may be input to the control calculation unit 154 and reflected in the calculation of the control variable (time ratio DTY).

[0034] In the example shown in Figure 2, the estimation unit 200 obtains the time ratio DTY from the input from the control calculation unit 154, but the estimation unit 200 may obtain the time ratio DTY in a different configuration. For example, it is also possible for the estimation unit 200 to obtain the time ratio DTY by arranging a time ratio measurement unit (not shown) after the PWM generation unit 156, and transmitting the measured value of the time ratio of (Ton / Ts) of the PWM signal ASpwn shown in Figure 2 from the time ratio measurement unit to the data holding unit 210.

[0035] (Details of circuit parameter estimation) Next, the parameter estimation according to this embodiment will be described in detail. Below, the operation of the estimation unit 200 for the main circuit 50A (step-down chopper) shown in Figure 2 will be described.

[0036] First, let's explain the equation of state. In this specification, the equation of state refers to an electrical relationship in which circuit parameters such as resistance, capacitance, and inductance are treated as constants, and voltage and current are treated as variables.

[0037] Figure 4 is a conceptual diagram illustrating an example of a state equation. Figure 4 illustrates an electrical circuit in which a current I flows when a voltage V is applied across a resistor R. The state equation for this electrical circuit is given by the following equation (1).

[0038] V = RI …(1) In electrical circuits involving switching, the time ratio of the switching elements is taken into consideration in the current-voltage relationship equation described above by the state averaging method described later. The circuit parameter estimation according to this embodiment is characterized in that the state equation and state averaging method, which are generally used to adjust the control gain during steady-state operation, are used to estimate parameters using data during transient operation.

[0039] Figure 5 is a circuit diagram illustrating the analysis model of the main circuit 50A (step-down chopper) shown in Figure 2.

[0040] As can be seen from the comparison of Figure 5 and Figure 2, the reactor 53 connected between nodes N1 and N2 has an inductance component (inductance value L) and a resistance component (resistance value R). Ｌ ) is represented by a series connection. Similarly, the capacitor 54 connected between nodes N2 and Ng has a capacitive component (capacitance value C) and a resistive component (resistance value R). Ｃ The equivalent circuit of the reactor and capacitor is shown as a series connection. The load 5 is represented by a resistor Ro (resistance value Ro) connected in parallel with the capacitor 54 between nodes N2 and Ng. The reactor current IL branches into a load current Io flowing through the load 5 and a current Ic flowing through the capacitor 54. A voltage Vc is generated in the capacitive component (capacitance value C) by charging and discharging with current Ic. In this embodiment, an example using the series connection circuit described above as the equivalent circuit of the reactor and capacitor will be explained below, but the equivalent circuit is not limited to this example and any configuration can be applied.

[0041] The analysis model in Figure 5 shows that the circuit state differs during the on-period and off-period of the switching element 51.

[0042] Figure 6A shows the equivalent circuit diagram of the ON period of the switching element 51, and Figure 6B shows the equivalent circuit diagram of the OFF period of the switching element 51.

[0043] When the switching element 51 shown in Figure 6A is turned on, the input voltage Vi from the power supply 2 is applied between nodes Ng and N1, which are connected in series and parallel to the reactor 53, capacitor 54, and load 5 (resistance value Ro), via the switched element 51 in the ON state, as indicated by the dotted line frame, resulting in a circuit state.

[0044] In contrast, when the switching element 51 shown in Figure 6B is turned off, the conduction diode 52, indicated by the dotted box, results in a circuit state where the potential difference between nodes Ng and N1 is approximately zero (input voltage Vi = 0).

[0045] In the circuit state shown in Figure 6A, the following circuit equations (2) to (4) hold true. Furthermore, by setting the left-hand side of equation (2) to 0 (Vi = 0), we can obtain the circuit equation shown in Figure 6B.

[0046]

[0047] Here, the reactor current IL and output voltage Vo are given by the state variables x(t) = [i Ｌ (t), v Ｏ (t) Ｔ Therefore, using equation (2) and the result of solving equation (4) for Vc and substituting it into equation (3), we obtain the state equation for Figure 6A (ON period of switching element 51) as a state coefficient matrix A, with each state coefficient as its component. １ , B １ We can obtain equation (5a) using this.

[0048] Also, in Equation (2), by setting Vi = 0, Equation (5b) using state coefficient matrices A2 and B2 can be obtained as the state equation for FIG. 6B (the off period of the switching element 51). Here, d[k] represents the duty ratio DTY in the k-th period. As described above, the period can be the switching period or the operation period. In the following description, it is assumed that sampling is performed for each switching period (period length Ts), and data for regression calculation is acquired.

[0049]

[0050] The state coefficient matrices A １ , A ２ , B １ , B ２ of each component (state coefficient) are shown by the following Equations (6) and (7).

[0051]

[0052] Hereinafter, in the present embodiment, an example in which the reactor current IL and the output voltage Vo are used as state variables will be described. However, any voltage or current included in the circuit equation can be arbitrarily used as the state variable. For example, in the state equation of the main circuit 50A (step-down chopper), it is also possible to use the voltage Vc of the capacitance component (capacitance value C) as the state variable. In this case, the voltage detector 72 can be arranged to detect the voltage (voltage between terminals) of the capacitor 54. That is, the voltage detector 72 can be arranged to detect at least one of the output voltage Vo or the voltage of the capacitor 54. Hereinafter, in the present embodiment, as shown in FIG. 2 and the like, the voltage detector 72 is arranged, so the output voltage detection value Voref corresponds to an example of the "voltage detection value". The voltage detector 72 corresponds to the "first voltage detector".

[0053] When the state equations (5a) and (5b) are averaged using the state averaging method by taking a weighted average using the time ratio of the switching element 51, the following equations (8) and (9) can be obtained as the state equations for the kth period (k: a natural number). Hereafter, the kth period (here, the switching period with a duration Ts) will also be simply referred to as "period k". In equation (9), d(kTs) represents the time ratio DTY at period k.

[0054]

[0055]

[0056] In equations (8) and (9), discretizing by setting x[k] = x(kTs) yields equation (10) below.

[0057]

[0058] Furthermore, equation (11) can be obtained by linearly approximating equation (10).

[0059]

[0060] Further rearranging equation (11) yields the following equation (12).

[0061]

[0062] Equation (12) shows that the difference between the state variable x with period k and the state variable x with period (k+1) can be calculated using the state coefficient matrices A[k] and B[k] and the state variable x[k] with period k. However, by changing our perspective on equation (12), we can interpret it as meaning that if we know the state variables x[k] and x[k+1] for periods k and (k+1), we can estimate the state coefficient matrices A[k] and B[k]. Based on this interpretation, we can derive equations (13) and (14) from equation (12) as regression equations where the state coefficient matrices A and B are treated as regression coefficients A(^) and B(^), thereby obtaining estimates of the state coefficient matrices A and B.

[0063] In this specification, matrices and parameter values ​​that are indicated with a "^" (hat) symbol in mathematical formulas to show estimated values ​​are written in the main text of the specification with a "(^)" symbol, as in A(^) above.

[0064]

[0065]

[0066] The regression coefficient in equation (14) is the state variable x[k] = [i] for N periods (N: a natural number greater than or equal to 2), based on the sampled output voltage detection value Voref and current detection value Iref. Ｌ (k), v Ｏ (k)] Ｔ This can be determined by performing a regression operation using the time ratio d[k]. This allows us to determine the state coefficient matrix A of the switching element 51 in the ON state. １ , B １ For each of the state coefficient matrices A2 and B2 in the off state, it is possible to calculate the estimated value of each component, i.e., each state coefficient. For example, the above estimated values ​​can be calculated using the least squares method, which is a general regression operation, according to equation (15) below.

[0067]

[0068] The state coefficient matrix A is obtained by the regression operation according to equation (15). １ , A ２ , B １ , B ２ Estimated state coefficient matrix A, which is composed of the estimated values ​​of each component. １ (^), A ２ (^) and estimated state coefficient matrix B １ (^), B ２ Each component of (^) can be expressed according to the following formulas (16a) and (16b).

[0069]

[0070] The estimated state coefficient matrix A shown in equations (16a) and (16b) １ (^), A ２ (^), B １ (^), B ２Each component of (^) and the state coefficient matrix A shown in equations (6) and (7) １ , A ２ , B １ , B ２ By comparing the coefficients of each component (state coefficient) including the circuit parameters, it is possible to pre-determine the estimation formula for each circuit parameter using the estimated value of any of the state variables.

[0071] Specifically, the inductance value L and the resistance value R of the resistive component of the reactor 53. Ｌ Furthermore, the capacitance value C of the capacitor 54 and the resistance value R of the resistive component. ｃ The respective estimated values ​​L(^) and R Ｌ (^)、C(^)、R Ｃ (^) can be determined from the estimated values ​​of each component shown in equations (16a) and (16b). That is, for the main circuit 50A which is composed of a step-down chopper, the circuit parameter estimation formula can be defined as the following equations (17a) to (17d).

[0072]

[0073] Note that the estimation formulas shown in equations (17a) to (17d) are just examples. For example, the estimated value L(^) of the inductance value L is given by the estimated state coefficient matrix A in equation (17a). １ The component (state coefficient) a １，１２ Although it is derived from the estimated values, the estimated state coefficient matrix A in equation (6) １ , A ２ Estimated value a of other components (state coefficients) that include the inductance value L inside １，１１ (^), a １，２１ (^), b １，１１ It is also possible to find it from (^).

[0074] The estimation calculation unit 220 sets a state variable x [k] based on the data (sampled detection values) acquired and held by the data holding unit 210, performs a regression calculation, and substitutes the estimated values ​​of the estimated state coefficient matrix obtained as a result of the regression calculation (equations (16a), (16b)) into the circuit parameter estimation equations (equations (17a) to (17d)) which are set in advance to correspond to the main circuit 50, thereby estimating the circuit parameters.

[0075] Furthermore, the timing of the calculation process for circuit parameter estimation by the estimation calculation unit 220 can be flexibly determined by providing a data holding unit 210. Specifically, the calculation of the estimated component values ​​of the state coefficient matrix in equations (16a) and (16b) by regression calculation using equation (15), and the substitution of the calculated component estimate values ​​into the circuit parameter estimation equation, does not necessarily have to be performed immediately in conjunction with the acquisition of the detected values ​​constituting the state variable x[k] by using data temporarily held in the data holding unit 210. In other words, the timing of the calculation process can be determined as appropriate.

[0076] For example, taking into account the processing speed of the arithmetic unit 100, the control unit 150 can prioritize the control calculations of the main circuit 50 (50A), while using the remaining processing power to divide and execute the calculations for circuit parameter estimation. This makes it possible to implement the circuit parameter estimation process according to this embodiment while suppressing the processing load on the arithmetic unit 100.

[0077] Furthermore, in Embodiment 1, the circuit parameter estimation formulas (equations (17a) to (17d)) are defined as the inductance value L and resistance value R of the reactor 53. Ｌ Furthermore, the capacitance value C and resistance value R of the capacitor 54. Ｃ We decided to calculate this, which corresponds to using the analysis models and equivalent circuits shown in Figures 5, 6A, and 6B.

[0078] For example, although omitted in this example for the sake of simplifying the equation, by modeling the resistance component of the switching element 51 and the resistance component of the diode 52, it becomes possible to include the resistance of the switching element 51 and the resistance of the diode 52 as targets for circuit parameter estimation. Alternatively, the forward voltage Vf of the diode 52 can also be modeled as a variable voltage source. In this case, by adding the resistance component of the switching element 51, the resistance component of the diode 52, and / or the variable voltage source of the output voltage Vf to the model in the areas indicated by the dotted lines in Figures 6A and 6B, circuit equations (equations (2) to (4)) can be created, thereby obtaining state equations (equations (5a) to (7)) and circuit parameter estimation equations that include these resistance components. As a result, the state coefficient matrix A can be obtained by the estimation processing calculation described above. １ , A ２ , B １ It is understood that by obtaining estimated values ​​for each component (state variable) of B2, it is also possible to calculate estimated values ​​for the resistance of the switching element 51, the resistance of the diode 52, and the forward voltage Vf of the diode 52.

[0079] Thus, even with the same main circuit 50 configuration, by modifying the analysis model and thereby modifying the state coefficient matrix and the circuit parameter estimation formula, the elements and their constant values ​​that are the target of circuit parameter estimation can be freely changed, as long as those constant values ​​are included in any of the components (state coefficients) of the state coefficient matrix that constitute the state equation. In other words, the circuit parameter estimation according to this disclosure can be freely determined using the concept of the state equation, provided that the state equation commonly used in the field of power electronics can be derived.

[0080] In Embodiment 1, circuit parameter estimation using the state equation was described when the step-down chopper is driven in continuous current mode or critical current mode. In contrast, in discontinuous current mode, a period of different states from the continuous current mode and critical current mode is added, specifically a period during which no reactor current flows. Therefore, the state equation and circuit parameter estimation formula differ from those of the continuous current mode and critical current mode. However, since a method for formulating the state equation has already been proposed for discontinuous current mode, the circuit parameter estimation according to this disclosure can also be applied to discontinuous current mode.

[0081] In this specification, in order to avoid unnecessarily complicating the explanation, which includes the embodiments described later, the circuit parameter estimation based on the equation of state in the current continuous mode or current critical mode will be explained.

[0082] Next, we will explain in detail the data held by the data holding unit 210, that is, the data used for regression calculations in the estimation calculation unit 220.

[0083] As described above, the data used in the regression calculation is periodic data defined to correspond to the switching period or the control calculation period. Since this data can generally be acquired at the same period as the data used to control the main circuit 50 (switching circuit), it can be easily acquired using the normal configuration and processing that the ADC 152 uses to sample the detected values ​​of the current detector 71 and voltage detector 72 for the control of the switching circuit. Therefore, the data used in the regression calculation for circuit parameter estimation can be acquired without requiring performance improvements to the current detector 71, voltage detector 72 and ADC 152, or additional processing by the ADC 152.

[0084] Furthermore, the regression calculation is performed using the difference between the value at period k and the value at period (k+1), as shown in equations (13) to (15). Therefore, if the power supplied to load 5 is DC (DC output), waveform data during transient operation (for example, transient operation) is required. This is because, in the case of DC output, the above difference hardly changes between periods during steady-state operation, making it difficult to secure enough data for an effective regression calculation.

[0085] Suitable transient operating periods for data acquisition in regression calculations include the startup (startup), shutdown (end of operation), load fluctuations, target value changes, or input voltage fluctuations of the power converter 10. Any transient operating period other than steady-state operation can be used. For example, from the listed phenomena, those that occur most frequently can be selected depending on the application of the power converter 10 to be used as the data acquisition period for regression calculations. It is desirable to use data from startup or shutdown, as the amplitude of change in the operating waveform of the switching circuit is considered to be large during these times.

[0086] Alternatively, a separate estimation mode can be provided to obtain a "data acquisition period" for regression calculations. In this estimation mode, by applying operations to the main circuit 50 (switching circuit) that change the time ratio and / or command value, data during non-steady-state operation can be intentionally acquired. In applications where non-steady-state operations such as startup or load fluctuations do not occur frequently, providing an estimation mode makes it possible to perform circuit parameter estimation at arbitrary timings (for example, at regular time intervals).

[0087] Furthermore, if the power supplied to load 5 is AC (AC output), data from steady-state operation can also be used in regression calculations for circuit parameter estimation. This is because, with AC output, changes occur in the difference of state variables for each switching cycle or control calculation cycle, even during steady-state operation. However, even with AC output, it is desirable to use data from non-steady-state operation, as utilizing data from multiple operating conditions is likely to improve estimation accuracy.

[0088] Furthermore, the state equation and circuit parameter estimation formula can be common regardless of whether the power supply to the load 5 is AC output or DC output. Generally, the period of the AC output is sufficiently long compared to the switching period or control calculation period of the switching element 51, so even with an AC output, the state equation can be derived by approximating it as a DC output.

[0089] Next, we will explain the number of data points (period number: N) used in the regression calculation. In principle, due to the structure of solving the equation, an indefinite number of data points (more than the number of circuit parameters to be derived) is required. However, since regression calculation is used, from the perspective of improving estimation accuracy, it is desirable that N be, for example, 10 times or more the number of circuit parameters to be estimated.

[0090] Furthermore, even if the calculation period is an integer multiple of the switching period Ts (Figure 3), the regression calculation in this embodiment can be performed by preparing the detected values ​​(Iref, Voref) and the average data of the time ratio DTY for one calculation period.

[0091] Furthermore, in this embodiment, while the circuit parameter estimation process using data acquired for each switching cycle or calculation cycle has been described as an example, the circuit parameter estimation process may be similarly performed using data acquired for multiple switching cycles or calculation cycles. However, assuming the duration of the transient operation described above is the same, the longer the data acquisition period, the fewer data points will be obtained for the regression calculation. For this reason, it is desirable that the data acquisition period (sampling period) for the regression calculation be the same as the switching cycle or calculation cycle.

[0092] As described above, in the power converter according to Embodiment 1, the data holding unit 210 of the estimation unit 200 shown in Figure 2 receives input data including the output voltage detection value and current detection value from the ADC 152 and the time ratio DTY from the control calculation unit 154. When DC output is supplied from the main circuit 50 (50A) to the load 5, the data holding unit 210 holds the input data for each switching cycle or control calculation cycle during non-steady operation, in the number of data points required for regression calculation. When AC output is supplied from the main circuit 50 (50A) to the load 5, the input data during steady operation may also be held in the data holding unit 210 for use in regression calculation.

[0093] The estimation calculation unit 220 of the estimation unit 200 can then perform a regression calculation using the data held in the data holding unit 210 to calculate estimated circuit parameter values ​​using a predetermined circuit parameter estimation formula corresponding to the main circuit 50 (switching circuit). In the example of Embodiment 1 described above, the circuit parameters of individual components mounted on the main circuit 50 (switching circuit) can be estimated by performing a regression calculation using input data of the output voltage detection value Voref, the current detection value Iref, and the time ratio DTY for each switching cycle or control calculation cycle. As can be understood from the above description, according to this disclosure, it is possible to estimate the circuit parameters of each component using easily obtainable detection values.

[0094] Embodiment 2. The circuit parameter estimation described in Embodiment 1 can also be applied to configurations other than the step-down chopper of the main circuit 50. Embodiment 2 describes examples of its application to various circuit configurations of the main circuit 50.

[0095] Figure 7 is a circuit diagram illustrating the analysis model of the main circuit 50B according to the first configuration example of Embodiment 2.

[0096] As shown in Figure 7, the main circuit 50B is composed of a boost chopper, and compared to the analysis model in Figure 5, the connection points of the switching element 51, diode 52, and reactor 53 are different.

[0097] In Figure 7, as in Figure 5, the reactor 53 has an inductance component (inductance value L) and a resistance component (resistance value R). Ｌ The capacitor 54 is shown as a series connection of ) and has a capacitive component (capacitance value C) and a resistive component (resistance value R Ｃ This is represented by a series connection of the following. Furthermore, the load 5 is represented by a resistor Ro (resistance value Ro) connected in parallel with the capacitor 54.

[0098] Therefore, circuit equations for the ON and OFF states of the switching element 51 can be established for the circuit configuration of the boost chopper. Furthermore, similar to the buck chopper (Embodiment 1), the state variable x(t) = [i Ｌ (t), v Ｏ (t) Ｔ As the state equations, equations (5a) and (5b) can be derived. Note that the derivation of the state equation for a boost chopper is publicly known, so the detailed process will not be described. In the state equation for a boost chopper (Figure 7), the state coefficient matrix A １ , A ２ , B １ , B ２ This is shown by equations (18) and (19) below.

[0099]

[0100] For the boost chopper as well, the regression equations are given by equations (13) and (14), and the state variable x[k] for N periods = [i Ｌ (k), v Ｏ (k)) Ｔ And by regression calculation using the time ratio d[k], the state coefficient matrix A of equations (18) and (19) is obtained. １ , A ２ , B １ , B ２ This allows us to obtain estimated values ​​for each component (state coefficient). This enables us to obtain the estimated state coefficient matrix A of equations (16a) and (16b), similar to Embodiment 1. １ (^), A ２ (^), B １ (^), B ２ Each component of (^) and the state coefficient matrix A of equations (18) and (19) １ , A ２ , B １, B ２ By comparing the coefficients of each component (state coefficient) including the circuit parameters, the following equations (20a) to (20d) can be obtained as circuit parameter estimation formulas for the main circuit 50B (boost chopper).

[0101]

[0102] As a result, the same circuit parameter estimation as in Embodiment 1 can be applied to the main circuit 50B, which is composed of a boost chopper, to calculate estimated values ​​of the circuit parameters of individual components mounted on the main circuit 50B (switching circuit).

[0103] Figure 8 is a circuit diagram illustrating the analysis model of the main circuit 50C according to the second configuration example of Embodiment 2.

[0104] As shown in Figure 8, the main circuit 50C is composed of a step-up / step-down chopper, and compared to the analysis model in Figure 5, the connection points of the switching element 51, diode 52, and reactor 53 are different.

[0105] In Figure 8, as in Figures 5 and 7, the reactor 53 has an inductance component (inductance value L) and a resistance component (resistance value R). Ｌ The capacitor 54 is shown as a series connection of ) and has a capacitive component (capacitance value C) and a resistive component (resistance value R Ｃ This is represented by a series connection of the following. Furthermore, the load 5 is represented by a resistor Ro (resistance value Ro) connected in parallel with the capacitor 54.

[0106] Therefore, circuit equations for the ON and OFF states of the switching element 51 can be established for the circuit configuration of the step-up / step-down chopper. Furthermore, similar to the step-down chopper (Embodiment 1), the state variable x(t) = [i Ｌ (t), v Ｏ (t) Ｔ As the state equations, equations (5a) and (5b) can be derived. Note that the derivation of the state equation for a buck-boost chopper is also publicly known, so the detailed process will not be described. In the state equation for a buck-boost chopper (Figure 8), the state coefficient matrix A １ , A ２ , B １ , B２ is represented by the following equations (21) and (22).

[0107]

[0108] For the buck-boost chopper as well, the regression equations are given by equations (13) and (14), and the estimated values of the components of the state coefficient matrices A Ｌ (k), v Ｏ (k)] Ｔ and the duty ratio d[k] can be obtained by regression calculation using the state variables x[k] = [i １ , A ２ , B １ , B ２ of equations (21) and (22). Further, similar to Embodiment 1, by comparing the coefficients of the estimated state coefficient matrices A １ (^), A ２ (^), B １ (^), B ２ (^) of equations (16a) and (16b) with the state coefficient matrices A １ , A ２ , B １ , B ２ of equations (21) and (22), the following equations (23a) to (23d) can be obtained as circuit parameter estimation equations for the main circuit 50C (buck-boost chopper).

[0109]

[0110] As a result, for the main circuit 50C composed of a buck-boost chopper as well, circuit parameter estimation similar to that in Embodiment 1 can be applied to calculate the estimated values of the circuit parameters of the individual components mounted on the main circuit 50C (switching circuit).

[0111] FIG. 9 is a circuit diagram for explaining an analysis model of the main circuit 50D according to the third configuration example of Embodiment 2. As will be described using FIG. 9, the circuit parameter estimation according to the present embodiment can also be applied to an insulated main circuit 50 (switching circuit) including a transformer.

[0112] As shown in FIG. 9, the main circuit 50D is constituted by a flyback converter including a switching element 51, a diode 52, a capacitor 54, and a transformer 55. The transformer 55 has a turns ratio of (1:n) between the primary side and the secondary side, and the primary side of the transformer 55 is connected to a power supply 2 constituted by a DC power supply via the switching element 51.

[0113] Furthermore, the secondary side of the transformer 55 is connected between nodes N1b and Ng2. Further, a diode 52, a capacitor 54, and a load 5 (resistance value Ro) are connected to the nodes N1b and Ng2 in the same configuration as that in FIG. 7 (boost chopper). Similar to FIGS. 5, 7, and 8, the capacitor 54 is represented by a series connection of a capacitance component (capacitance value C) and a resistance component (resistance value R Ｃ ).

[0114] FIG. 10 shows an equivalent circuit diagram of the main circuit 50D shown in FIG. 9. As shown in FIG. 10, the transformer 55 is equivalently converted into a series connection circuit of a resistance component (resistance value Rtr) and an inductance component (inductance value Ltr). Thereby, the capacitance value C♯, the resistance value Rc♯ of the capacitor 54 on the secondary side, and the resistance value Ro♯ of the load 5 are also converted to the primary side (Rc♯ = Rc / n, Rc♯ = Rc, Ro♯ = Ro / n).

[0115] Note that, regarding the equivalent circuit of the transformer 55, in addition to the example shown in FIG. 10, it may be modeled by a configuration in which leakage inductance is separately arranged on the primary side and the secondary side, or it may be modeled to include stray capacitance.

[0116] Regarding the flyback converter exemplified as an isolated switching circuit, based on the equivalent circuit of FIG. 10, circuit equations for the on period and the off period of the switching element 51 can be established. Thereby, although detailed description is omitted, using the transformer current Itr flowing through the transformer 55 and the output voltage Vo as state variables, state equations (Equations (5a) and (5b)) can be derived in the same manner as described in the first and second examples of Embodiment 1 and Embodiment 2. The state coefficient matrix A １ , A ２ , B１ , B ２ Each component can be appropriately determined from the circuit equations described above.

[0117] Therefore, for the main circuit 50D (flyback converter) as well, the regression equations (13) and (14) are given by the state variable x [k] = [i ｔｒ (k), v Ｏ (k)] Ｔ And by performing a regression operation using the time ratio d[k], the state coefficient matrix A １ , A ２ , B １ , B ２ The estimated values ​​of each component can be obtained. Specifically, similar to Embodiment 1, the estimated state coefficient matrix A of equations (16a) and (16b) can be obtained. １ (^), A ２ (^), B １ (^), B ２ Each component of (^) and the state coefficient matrix A of equations (18) and (19) １ , A ２ , B １ , B ２ Each component can be compared. Therefore, the circuit parameter estimation formula for the main circuit 50D (flyback converter) can also be predetermined in the same way as for the main circuits 50A to 50C.

[0118] As a result, the same circuit parameter estimation as in Embodiment 1 can be applied to the flyback converter (main circuit 50D), which is an isolated switching circuit, to calculate estimated values ​​of the circuit parameters of individual components mounted on the main circuit 50D.

[0119] The circuit parameter estimation method according to this embodiment is also applicable to power conversion devices whose main circuit is a switching circuit using multiple switching elements.

[0120] Figure 11 is a circuit diagram illustrating the analysis model of the main circuit 50E according to the fourth configuration example of Embodiment 2.

[0121] Referring to Figure 11, the main circuit 50E includes a full-bridge circuit 51x consisting of switching elements Q1 to Q4, a reactor 53, and a capacitor 54.

[0122] The reactor 53 and capacitor 54 are connected in series via node N2 between node Na, which corresponds to the connection point of switching elements Q1 and Q2, and node Nb, which also corresponds to the connection point of switching elements Q1 and Q2. The load 5 (resistance value Ro) is connected in parallel with the capacitor 54, between node N2 and node Nb.

[0123] In Figure 11, as in previous examples, the reactor 53 has an inductance component (inductance value L) and a resistance component (resistance value R). Ｌ The capacitor 54 is shown as a series connection of ) and has a capacitive component (capacitance value C) and a resistive component (resistance value R Ｃ This is represented by a series connection of ).

[0124] When the main circuit includes multiple switching elements, the state equation can be derived by setting up circuit equations for all switching patterns, which are combinations of on / off states of those multiple switching elements.

[0125] Furthermore, by averaging the state equations obtained for each switching pattern using the time ratio (state averaging method), and applying the regression operation using equation (15), etc., to the regression equation obtained from the averaged state equations, it is possible to obtain estimated values ​​for each component (state coefficient) of the state coefficient matrix that constitutes the state equation, as in the previous examples. In addition, the circuit parameter estimation formula for the main circuit 50E can also be predetermined in the same way as for the main circuits 50A to 50D.

[0126] For example, if the full-bridge circuit 51x is operated so that the set of switching elements Q1 and Q4 and the set of switching elements Q2 and Q3 are turned on alternately, the same circuit parameter estimation formula as the main circuit 50A (buck chopper) in Embodiment 1 can be obtained.

[0127] Furthermore, the circuit parameter estimation method according to this embodiment is also applicable to multilevel circuits such as cascade inverters or multiphase circuits such as interleaved circuits.

[0128] Figure 12 is a circuit diagram illustrating an analysis model of the main circuit 50F according to a fifth configuration example of Embodiment 2. The main circuit 50F is composed of a two-phase interleaved boost converter.

[0129] Referring to Figure 12, the main circuit 50F includes switching elements Q1 and Q2, reactors 53-1 and 53-2, and a capacitor 54.

[0130] Reactor 53-1 is connected between nodes N1 and Nc, and the inductance component (inductance value L) １ ) and the resistive component (resistance value R Ｌ1 This is shown as a series connection of ). Reactor 53-2 is connected between nodes N1 and Nd, and the inductance component (inductance value L) ２ ) and the resistive component (resistance value R Ｌ2 This is represented by a series connection of ).

[0131] Switching element Q1 is connected between nodes Nc and Ng, and switching element Q2 is connected between nodes Nd and Ng. Switching elements Q1 and Q2 interleave by being controlled on and off in inverted phase according to the time ratio DTY by separate control signals.

[0132] Diode 52-1 is connected between nodes Nc and N2, and diode 52-2 is connected between nodes Nd and N2. Capacitor 54 is connected between nodes N2 and Ng and is represented by a series connection of a capacitive component (capacitance value C) and a resistive component (resistance value Rc).

[0133] As a result, the main circuit 50F has a two-phase interleaved boost converter circuit configuration in which boost converters 61 and 62 are connected in parallel between nodes N1 and N2. Boost converter 61 is composed of a switching element Q1, a diode 52-1, and a reactor 53-1. Similarly, boost converter 62 is composed of a switching element Q2, a diode 52-2, and a reactor 53-2.

[0134] In the main circuit 50F, the reactor current IL1 of reactor 53-1 of the first-phase boost converter 61, the reactor current IL2 of reactor 53-2 of the second-phase boost converter 62, and the output voltage Vo can be detected and expressed as a state variable x(t). That is, x(t) = [i Ｌ１ (t), i Ｌ２ (t), v Ｏ (t) Ｔ We can derive the following equation of state.

[0135] Although the main circuit 50F is illustrated as an interleaved circuit with "2" phases, it is also possible to use three or more phases. In this case, the state equations and circuit parameter estimation formulas described later can be obtained by changing the number of reactor current detections and the dimensions of the state variables and state equations according to the number of phases.

[0136] In the main circuit 50F (two-phase interleaved boost converter) shown in Figure 12, switching elements Q1 and Q2 are switched on and off by inverting their phases. Therefore, different circuit states exist depending on the time ratio DTY of switching elements Q1 and Q2.

[0137] Specifically, when the time ratio DTY ≤ 0.5, there are three possible on / off combinations (i.e., switching patterns) for the switching elements Q1 and Q2: (Q1, Q2) = (on, off), (off, on), and (off, off). On the other hand, when the time ratio DTY > 0.5, there are also three possible on / off combinations (i.e., switching patterns) for the switching elements Q1 and Q2: (Q1, Q2) = (on, off), (off, on), and (on, on).

[0138] Thus, since the circuit states (switching patterns) that exist differ depending on the time ratio DTY, different state equations and circuit parameter estimation formulas are derived depending on the time ratio DTY. In other words, it is necessary to pre-determine the circuit parameter estimation formulas for each of the two cases, time ratio DTY ≤ 0.5 and DTY > 0.5, and switch the circuit parameter estimation formula according to the time ratio DTY at each point in time. As described above, in this embodiment, as explained in Figure 2, the value of the time ratio DTY is sequentially input to the data holding unit 210 and the estimation calculation unit 220, so it is possible to switch the circuit parameter estimation formula according to the time ratio DTY.

[0139] Below, we will illustrate the state equation and circuit parameter estimation formula for the case where DTY > 0.5, one of the two cases mentioned above.

[0140] Figure 13 is an example of an operating waveform diagram of the main circuit 50F shown in Figure 12. Figure 13 shows an example of operation when DTY > 0.5.

[0141] Referring to Figure 13, the period from time tx to tz corresponds to period k, and from time tz, period (k+1) begins. Time ty is the midpoint between times tx and tx, and is the timing when half a period (Ts / 2) has elapsed from time tx.

[0142] In period k, the ON period of switching element Q1 starts at time tx and ends at time tb, which is the time elapsed from time tx by the multiplication value (d1・Ts) of the time ratio d1 (d1 = DTY) and the period Ts in the boost converter 61. On the other hand, the ON period of switching element Q2 starts at time ty and ends at time tc, which is the time elapsed from time ty by the multiplication value (d2・Ts) of the time ratio d2 (d2 = DTY) and the period Ts in the boost converter 62.

[0143] Since DTY > 0.5, the time tb at which switching element Q1 turns off is later than time ty, and the time tc at which switching element Q2 turns off is later than the time tz at which period (k+1) begins. Also, the time ta at which the on period of switching element Q2 from the previous period (k-1) ends is later than the time tx at which period k begins.

[0144] The reactor current IL1 increases during the ON period of switching element Q1 and decreases during the OFF period. Similarly, the reactor current IL2 increases during the ON period of switching element Q2 and decreases during the OFF period.

[0145] One period Ts is divided into periods α1 to α4 in which the switching elements Q1 and Q2 change. In periods α1 and α3, (Q1, Q2) = (on, on), in period α2, (Q1, Q2) = (on, off), and in period α4, (Q1, Q2) = (off, on).

[0146] The equation of state can be expressed as follows for each of the periods α1 to α4: equations (24a) to (24d).

[0147]

[0148] State coefficient matrix A in equations (24a) and (24c) １ , B １ and A ３ , B ３ This is shown by equation (25) below.

[0149]

[0150] Similarly, the state coefficient matrix A in equation (24b) ２ , B ２ This is expressed by the following equations (26a) and (26b).

[0151]

[0152] Also, the state coefficient matrix A in equation (24d) ４ , B ４ This is shown by the following equations (27a) and (27b).

[0153]

[0154] Then, by averaging the state equations obtained for each switching pattern as described above (state averaging method), and applying the regression operation using equation (15), etc., to the regression equation obtained based on the averaged state equations, it is possible to obtain estimated values ​​for each component of the state coefficient matrix that constitutes the state equation, as in the previous examples. Furthermore, regarding the circuit parameter estimation equation for the main circuit 50F, equations (28a) to (28d) below can be obtained from the components of the state coefficient matrix.

[0155]

[0156] As a result, the circuit parameter estimation according to this embodiment can also be applied to the main circuit 50F, which is composed of a two-phase interleaved boost converter, to calculate estimated values ​​of the circuit parameters of individual components mounted on the main circuit 50F (switching circuit).

[0157] Thus, even in multilevel or multiphase circuits, detection values ​​to be used as state variables can be set according to the configuration of the main circuit 50, and a state equation can be formulated for each switching pattern of multiple switching elements by performing an operation analysis. Furthermore, by averaging these state equations (state averaging method) and applying a regression operation to the regression equation derived based on the averaged state equation, it is possible to obtain estimated values ​​for each component of the state coefficient matrix that constitutes the state equation. As a result, even for the main circuit 50 configured as a multilevel or multiphase circuit, the circuit parameter estimation method according to this embodiment can be applied by similarly obtaining a circuit parameter estimation formula based on each component of the state coefficient matrix.

[0158] Furthermore, it should be noted for clarification that, in circuit parameter estimation for a main circuit whose circuit state switches according to a time ratio, as illustrated in Figure 12, it is not necessarily required to pre-determine the circuit parameter estimation formulas for all circuit states. Specifically, by preparing only the circuit estimation parameters corresponding to some of the circuit states, it is possible to achieve similar circuit parameter estimation by using only the detected values ​​for those specific circuit states as input data to the data holding unit 210 for regression calculations.

[0159] Furthermore, in embodiments 1 and 2, a resistive load (resistance value Ro) was exemplified as the load 5. However, even if the load 5 receiving power from the main circuit 50 is not a resistive load, the circuit parameter estimation according to this embodiment can be applied. For example, the load 5 may be composed of an inductive L load or a capacitive C load, or at least two of R (resistance), L, and C are connected in series, parallel, or series-parallel. Alternatively, the load 5 may be equivalently composed of a voltage source or a current source, or may be set by a motor or power system. In any of these cases, if it is possible to derive a state equation from a circuit equation including the load 5 and the main circuit 50, treating the load 5 as an element in the electrical circuit, then it is possible to implement the circuit parameter estimation according to this embodiment for the main circuit 50 that supplies power to the load 5.

[0160] Embodiment 3. Embodiment 3 describes a technique for improving the accuracy of circuit parameter estimation by using the average value of the data used in regression calculations within the above-mentioned period (a period corresponding to the switching period or calculation period).

[0161] As described above, in the circuit parameter estimation for this implementation, a state variable x[k] = [i] is calculated for N periods (N: a natural number of 2 or more) based on a voltage detector (e.g., output voltage detection value Voref) and a current detection value Iref, which are sampled at each period. Ｌ (k), v Ｏ (k)] Ｔ The time ratio d[k] is stored in the data holding unit 210 and used in the regression calculation.

[0162] In this case, the state variable x[k] = [i] used in the regression calculation Ｌ (k), v Ｏ (k)] Ｔ Furthermore, by using the average value data within each period instead of simple sample values ​​for each period for the time ratio d[k], the accuracy of circuit parameter estimation can be improved.

[0163] Figure 14 shows an example configuration of an average value calculation circuit 90 for obtaining the average value within a period.

[0164] As shown in Figure 14, the average value calculation circuit 90 according to Embodiment 3 includes a reset switch 91, a transconductance amplifier 92, and a capacitor 93. The transconductance amplifier 92 is configured to output a current corresponding to the input voltage Vdet to node Nz. The capacitor 93 is connected to node Nz and is charged by the output current of the transconductance amplifier 92.

[0165] The reset switch 91 is typically composed of a transistor and turns on in response to the reset signal Rst. When the reset switch 91 is turned on, the capacitor 93 is discharged, and the voltage at node Nz is initialized to 0 (GND). The reset signal Rst is generated so that the reset switch 91 turns on at each period in which the regression calculation data is sampled, for example, at each switching period or calculation period.

[0166] In the average value calculation circuit 90, node Nz generates an average voltage Vave, which corresponds to the integral value of the input voltage Vdet for each period in which the reset switch 91 is turned on. Therefore, by setting the input voltage Vdet to the output voltage of the current detector 71 or the voltage detector 72, the average value for each period of the reactor current IL (IL1, IL2) or output voltage Vo can be obtained as the average voltage Vave.

[0167] Therefore, in the configuration shown in Figure 2, the ADC 152 can convert the output voltage (average voltage Vave) of the average value calculation circuit 90 in Figure 14 into a digital value and output it as the output voltage detection value Voref and the current detection value Iref. As a result, the data holding unit 210 can hold the average value data within each period as state variables (reactor current IL (IL1, IL2) and output voltage Vo) used in the regression calculation. This improves the accuracy of circuit parameter estimation by the regression calculation.

[0168] Alternatively, similar average value data can be obtained by shortening the sampling period at the ADC 152 and sampling the output of the current detector 71 or voltage detector 72 multiple times within a single sampling or calculation period. In this case, for example, a function can be added to the estimation unit 200 to calculate the average value of the digital values ​​obtained from multiple samplings at the ADC 152 for each sampling or calculation period.

[0169] Alternatively, by setting the sampling timing using the ADC152, it is possible to obtain a voltage or current equivalent to the average value.

[0170] Figure 15 is a waveform diagram illustrating the sampling timing of the ADC152 according to Embodiment 3.

[0171] Figure 15 shows a waveform diagram of PWM control similar to that in Figure 3. According to the comparison of the calculated values ​​of the carrier wave CW and the time ratio DTY, time t1 to t2 is the ON period for the switching element 51, etc., and time t2 to t3 is the OFF period for the switching element 51, etc.

[0172] As shown in Figure 13, the reactor current IL(IL1, IL2) generally increases at a constant rate during the ON period of the switching element and decreases at a constant rate during the OFF period. The output voltage Vo also changes in conjunction with the reactor current IL(IL1, IL2).

[0173] Therefore, by setting the sampling timing to time ts1, which corresponds to the midpoint of the ON period (times t1 to t2) of the switching element 51, etc., or time ts2, which corresponds to the midpoint of the OFF period (times t2 to t3), the sampled values ​​of the reactor current IL (IL1, IL2) and output voltage Vo will be close to their respective average values.

[0174] The optimal sampling times ts1 and ts2 correspond to the peaks or troughs of a triangular wave when the carrier wave CW is a triangular wave.

[0175] By setting the sampling timing of the ADC 152 at time ts1 or ts2 in Figure 15, the output voltage detection value Voref and current detection value Iref can be made equivalent to the average value for each period by sampling once per period. This improves the accuracy of circuit parameter estimation by using the average value data for each period in regression calculations without adding the circuits or processes described in Figure 14, etc.

[0176] Furthermore, it is possible to calculate average values ​​using AI (Artificial Intelligence) technologies such as neural networks.

[0177] Figure 16 shows an example of the configuration of the neural network model 80 according to Embodiment 3.

[0178] Referring to Figure 16, the neural network model 80 includes K neurons NL11 to NL1K (in the example in Figure 16, K: an integer greater than or equal to 3) that constitute the input layer, I neurons NL2 (I: a natural number, K ≥ I ≥ 1) that constitute the output layer, and a plurality of neurons that constitute a hidden layer connected between the input and output layers. Furthermore, the hidden layer is composed of up to M neurons (M: an integer greater than or equal to 2) interconnected across J layers (J: a natural number, J ≥ 1). The structure of the neural network model 80 is set by determining the above-mentioned numerical parameters K, I, J, and M.

[0179] The structure of the neural network model 80 can be arbitrarily configured by the number of input layers, hidden layers, and output layers, as well as the number of neurons in each layer. Each neuron, indicated by a circle symbol in Figure 16, is input with an activation function. For example, a sigmoid function can be used as the activation function, but any known activation function can be applied.

[0180] The input layer receives multiple (three or more) detected values ​​(sampling values) of current (e.g., reactor current IL) or voltage (e.g., output voltage) at consecutive sampling timings within the same period or multiple consecutive periods, using at least three neurons. The sampling timing of these multiple detected values ​​can be predetermined to correspond to the phase of the carrier wave, so that sampling occurs once or multiple times (i.e., one or more times) in each period (carrier period).

[0181] In the output layer, neuron NL2 generates average value data for each period between 1 and K corresponding to at least some of the multiple (K) sampled values ​​from the input layer. That is, the multiple data are not limited to data within the same period as described above, but may include sampled values ​​from the periods before and after the period for which the average value is calculated. In the example in Figure 16, K ≥ 3 and I = 1, but as described above, I can be set to K if the K sampled values ​​are sampled from different periods, so as described above, 1 ≤ I ≤ K.

[0182] The weight coefficients between each neuron in the neural network model 80 can be determined by machine learning. For example, the weight coefficients can be set by machine learning that takes the above-mentioned multiple sample values ​​as input values ​​for the actual reactor current or output voltage, and the actual average value at that time as the output value (training data).

[0183] A neural network model 80 with weight coefficients determined by such machine learning can be realized by software processing by the estimation unit 200. Furthermore, the sampling timing of the reactor current (current detector 71) or output voltage Vo (voltage detector 72) by the ADC 152 can be set to a timing corresponding to the input values ​​of the machine learning.

[0184] This makes it possible to input multiple sampled values ​​from the ADC152 into the neural network model 80, thereby obtaining the average value (estimated value) of the current or voltage for each period as the output value of the neural network model 80. In particular, by inputting sampled values ​​of current or voltage within multiple periods, it is possible to obtain average value data for each period using one sampled value for each carrier wave period Ts (switching period Ts) without shortening the sampling period.

[0185] Furthermore, in order to improve the learning accuracy of the neural network model 80, it is also possible to add data other than the detected value to the input value of the input layer, such as the output voltage command value, input voltage detected value, time ratio, switching frequency, etc.

[0186] As described in Embodiment 3, by using average value data within one period obtained by any method as data for regression calculation in the estimation calculation unit 220, the accuracy of circuit parameter estimation can be improved.

[0187] Embodiment 4. Embodiments 1 to 3 described an example in which the power supply 2 connected to the main circuit 50 is a DC power supply. Embodiment 4 describes an example in which the power supply 2 is an AC power supply. Even if the power supply 2 is an AC power supply, the state equation can be derived, and the circuit parameter estimation according to this embodiment can be applied. However, in this case, the input voltage Vi also needs to be detected as a variable.

[0188] Figure 17 shows the first configuration of the power converter according to Embodiment 4. Comparing Figure 17 and Figure 2, the power converter according to Embodiment 4 differs from the power converter shown in Figure 2 (Embodiment 1) in that the power supply 2 is configured as an AC power supply and a rectifier circuit 20 is further arranged. The rectifier circuit 20 can typically be configured as a diode bridge. Also, for the sake of simplicity, the main circuit 50 (50A) is configured as a step-down chopper, similar to Figure 2.

[0189] As a result, in Figure 17, the input voltage Vi to the main circuit 50A is not a constant voltage from the DC power supply, but a rectified AC voltage that changes to have a ripple component according to the frequency of the AC voltage. Therefore, a voltage detector 73 is further provided in the main circuit 50A to detect the input voltage Vi. The reactor current IL and output voltage Vo are detected by the current detector 71 and the voltage detector 72, as in Figure 2. The voltage detector 73 corresponds to the "second voltage detector".

[0190] In Embodiment 4, the control unit 150, in addition to the reactor current IL and output voltage Vo, converts the input voltage Vi (analog value) detected by the voltage detector 73 into a digital value at a predetermined period and outputs the output voltage detection value Voref, the current detection value Iref, and the input voltage detection value Viref.

[0191] The control calculation unit 154 receives an input voltage detection value Viref in addition to the output voltage detection value Voref from the ADC 152, and can calculate the time ratio DTY of the switching element 51 for controlling the output voltage Vo. For example, the control calculation unit 154 may calculate the time ratio DTY by a combination of feedback control based on the output voltage detection value Voref, similar to Embodiment 1, and feedforward control based on the input voltage detection value Viref.

[0192] The data holding unit 210 receives an input voltage detection value Viref from the ADC 152, in addition to the output voltage detection value Voref, current detection value Iref, and time ratio DTY, similar to those in Embodiment 1, and stores this information (data for regression calculation) necessary for circuit parameter estimation.

[0193] As mentioned above, when an AC voltage is input from power supply 2, even during steady-state operation, the difference between periods in the data (sampling values) of the reactor current IL and output voltage Vo for each switching cycle or control calculation cycle may change. Therefore, data from steady-state operation can also be used for regression calculations in the estimation calculation unit 220. However, as with the case where power supply 2 is a DC power supply, data from transient operations such as startup or shutdown is preferable for regression calculations.

[0194] The estimation calculation unit 220 can perform a regression calculation using the data including the input voltage detection value Viref held in the data holding unit 210, and calculate estimated values ​​of the circuit parameters using the circuit parameter estimation formula prepared for the main circuit 50A, similar to the first embodiment.

[0195] Furthermore, regarding the equations of state and regression equations, the same formulas can be used as when the input voltage Vi is a DC voltage by changing the input voltage Vi from a fixed value (constant) to a variable (measured value). Specifically, by changing the input voltage Vi, which is treated as a constant in the formulas described in Embodiment 1, etc., to the input voltage Vi(k) at period k, it is possible to handle the case where the power supply 2 is an AC voltage. The sampled input voltage detection value Viref can be substituted for the input voltage Vi(k).

[0196] In the embodiment 1, where the power supply 2 is configured as a DC power supply, a voltage detector 73 for the input voltage Vi can be further provided, allowing information on the input voltage Vi to be used for circuit parameter estimation. This is because, even when the power supply 2 is a DC power supply, the input voltage Vi may fluctuate in accordance with the power conversion operation of the main circuit 50 (50A) due to the responsiveness of the power supply 2 or the influence of wiring impedance, etc. Whether the power supply 2 is a DC power supply or an AC power supply, the accuracy of circuit parameter estimation can be improved by using information (detected value) of the input voltage Vi of the main circuit 50 for circuit parameter estimation.

[0197] Although Figure 17 illustrates an example configuration in which the input voltage Vin of the main circuit 50 is generated using a rectifier circuit 20 (diode bridge), the circuit parameter estimation according to this embodiment can also be applied to configurations in which an AC voltage is directly input to the main circuit 50 without using a diode bridge.

[0198] Figure 18 is a circuit diagram illustrating the configuration of the main circuit 50X in a second configuration example of the power conversion device according to Embodiment 4.

[0199] The main circuit 50X shown in Figure 18 includes switching elements Q1 and Q2, diodes 52-1 and 52-2, and reactors 53-1 and 53-2, which constitute a so-called bridgeless boost ADC converter.

[0200] In the main circuit 50X, two boost converters 61 and 62 are connected to node N2, to which capacitor 54 and load 5 are connected. Boost converter 61 consists of a switching element Q1, a diode 52-1, and a reactor 53-1. Similarly, boost converter 62 consists of a switching element Q2, a diode 52-2, and a reactor 53-2.

[0201] During periods when the input voltage Vi > 0, switching element Q2 is kept ON while switching element Q1 is controlled ON / OFF, causing boost converter 61 to perform ADC conversion with boost operation. On the other hand, during periods when the input voltage Vi < 0, switching element Q1 is kept ON while switching element Q2 is controlled ON / OFF, causing boost converter 62 to perform ADC conversion with boost operation.

[0202] Although a detailed explanation is omitted, a state equation can also be derived for each switching pattern (on / off combination) of switching elements Q1 and Q2 in the main circuit 50X (bridgeless boost ADC converter). Therefore, using the same method as in Embodiment 1, the state equation for each switching pattern can be averaged using the time ratio of switching elements Q1 and Q2 (state averaging method), and the circuit parameter estimation according to this embodiment can be applied by deriving a regression equation based on the averaged state equation and a circuit parameter estimation equation based on the components of the state coefficient matrix.

[0203] In Figures 11 and 18, single-phase circuits are shown as examples when the load 5 is an AC circuit or when the power supply 2 is an AC power supply. However, the state equation can also be derived for three-phase AC circuits (three-phase AC / DC converters and three-phase DC / AC converters). Therefore, by deriving the regression equation and circuit parameter estimation equation based on the state equation in the same manner as in Embodiments 1 to 4, the circuit parameter estimation according to this embodiment can also be applied when the main circuit 50 is a three-phase AC circuit.

[0204] Furthermore, in Embodiment 4, the load 5 may be configured as an inductance-only L load or a capacitance-only C load, as described in Embodiments 1 and 2, or by connecting at least two of R (resistance), L, and C in series, parallel, or series-parallel. Alternatively, the load 5 may be equivalently configured as a voltage source or a current source, or it may be set by a motor or power system. That is, if the load 5 is treated as an element in an electrical circuit, and it is possible to derive a state equation from a circuit equation including the load 5 and the main circuit 50, then it is possible to implement circuit parameter estimation according to this embodiment for the main circuit 50 that supplies power to the load 5.

[0205] As described in Embodiments 1 to 4 above, according to this embodiment, in a power conversion device that supplies power from a power source 2 to a load 5 by power conversion using a main circuit comprising one or more reactors or transformers, one or more switching elements, and one or more capacitors, the circuit parameters of the components constituting the main circuit can be estimated. More specifically, estimated values ​​of the circuit parameters can be calculated by substituting the regression calculation results using the sampling values ​​and time ratios of current and voltage acquired during the "data acquisition period," which includes non-steady-state operation, into a predetermined circuit parameter estimation formula corresponding to the circuit topology of the main circuit.

[0206] In this case, if the regression equation and circuit parameter estimation equation used in the regression calculation can be derived from the state equation derived in accordance with the circuit topology of the main circuit, then the circuit parameters included in the components of the state coefficient matrix of the state equation can be estimated without limiting the circuit topology. As explained above, the estimated circuit parameters can include at least one of the following: resistance, capacitance, inductance, and forward voltage (diode). Furthermore, for the equivalent circuits of capacitors and reactors, any circuit configuration can be adopted as long as the state equation can be derived in combination with the circuit topology of the main circuit.

[0207] Furthermore, the data used in the regression calculation may include multiple periods of data within the data acquisition period (each switching period or calculation period) for the current flowing through one or more reactors or transformers, the voltage such as the output voltage to the load, and the time ratio of one or more switching elements. This data can be acquired during the "data acquisition period," which includes transient operations such as startup, shutdown, load fluctuations, target value changes, input voltage fluctuations, or estimation operation modes of the power converter. In addition, the estimation accuracy of circuit parameters can be improved by using the average value of each period for the periodic data used in the regression calculation.

[0208] Furthermore, in this embodiment, circuit parameter estimation can be performed by substituting the result of a regression calculation using a pre-prepared regression equation and the data stored in the data holding unit 210 into a pre-prepared regression parameter equation. Therefore, after the data for the regression calculation is stored in the data holding unit 210, the timing of the circuit parameter estimation process is not particularly restricted. As a result, the timing of the estimation process can be flexibly adjusted by lowering its priority than the control calculation of the main circuit 50, so that the estimated circuit parameter values ​​can be calculated without excessively increasing the processing load of the arithmetic unit 100.

[0209] (Use of Estimated Circuit Parameter Values) In this embodiment, as explained in Figure 2, the estimation unit 200 can transmit the estimated circuit parameter information (typically, the estimated values ​​of the circuit parameters) to the control unit 150 and / or the power converter 10 either as is or after signal processing. As will be explained below, the estimated circuit parameters can be used to generate abnormal alarms and to optimize the operation of the power converter 10 (main circuit 50) by changing the circuit operation and adjusting the control parameters, either through processing inside or outside the arithmetic unit 100.

[0210] As a first example, the estimated circuit parameters can be used to perform anomaly detection or degradation judgment. Specifically, a threshold value for a certain circuit parameter can be set in advance, and if the estimated value rises or falls below the threshold value, an anomaly can be detected, and an anomaly alarm can be output to the outside of the power converter 10 (arithmetic unit 100). The anomaly alarm corresponds to one embodiment of the "signal based on the estimated value of the circuit parameter".

[0211] For example, an abnormal alarm related to a switching element (MOSFET) can be output when its resistance (estimated value) rises above a predetermined threshold. Alternatively, an abnormal alarm related to a switching element can be output when the capacitance (estimated value) of the capacitor 54 falls below a predetermined threshold. Furthermore, the temperature may be estimated from the estimated forward voltage of the diode, and an abnormal alarm warning against leaving the device at high temperature may be output when the estimated temperature rises above a predetermined threshold.

[0212] Furthermore, the abnormal alarm may be generated outside the power converter 10 (arithmetic unit 100) based on the estimated circuit parameter values ​​transmitted from the estimation unit 200. In other words, abnormal diagnosis may be performed outside the power converter 10 (arithmetic unit 100).

[0213] In contrast, the control unit 150 (Figure 2) may directly detect an abnormality from the circuit parameter estimates from the estimation unit 200 and control the power converter 10 to stop the operation of the main circuit 50 when an abnormality is detected. For example, the PWM generation unit 156 can stop the power conversion operation of the main circuit 50 by fixing the PWM signal Spwm to an L level so as to turn off the switching element 51. Alternatively, the PWM generation unit 156 may turn off the switching element 51 in response to an abnormality alarm.

[0214] As a second example, the estimated circuit parameters can be used to change the circuit operation of the power converter 10 (main circuit 50).

[0215] For example, as the capacitance of a capacitor decreases, the voltage ripple increases. In response to this, if the estimated capacitance of the capacitor decreases within a range higher than the threshold for anomaly detection, the switching frequency of the switching element (the reciprocal of the switching period Ts) can be increased to suppress the voltage ripple. Also, to prevent the output voltage Vo from exceeding the rated voltage of the capacitor, the target value of the output voltage Vo can be lowered in response to the decrease in the estimated capacitance of the capacitor.

[0216] On the other hand, as the driving frequency of the switching element increases, the switching loss increases, so the switching frequency can be changed in response to the change in the capacitance value C (estimated value) of the capacitor. Specifically, the smaller the capacitance (estimated value), the higher the switching frequency can be set, while the larger the capacitance (estimated value), the lower the switching frequency can be set. The adjustment of the switching frequency is achieved by changing the period of the carrier wave CW (Figure 3) etc. in the PWM generation unit 156.

[0217] Furthermore, if the inductance value L of the reactor decreases, the current ripple will increase. Therefore, the current limit can be modified to restrict the upper limit of the maximum current so that the reactor current IL (IL1, IL2) does not exceed the reactor's saturation current. For example, the switching frequency of the switching element can be changed to suppress the increase or decrease of current ripple in response to the change in the inductance value L (estimated value). Specifically, the smaller the inductance (estimated value), the higher the switching frequency can be set, while the larger the inductance (estimated value), the lower the switching frequency can be set. Alternatively, in response to the decrease in inductance (estimated value), it is also possible to lower the target value of the output voltage Vo in order to lower the average value of the reactor current.

[0218] Alternatively, to suppress the maximum current, a protection function can be incorporated into the PWM generation unit 156 that forcibly turns off the switching element by forcibly setting the PWM signal Spwm to L level when the detected value of the reactor current IL exceeds the maximum current. This protection function can then be turned on when the inductance value L (estimated value) of the reactor decreases. Alternatively, when the resistance (estimated value) of the switching element, the resistance (estimated value) of the diode, and the forward voltage (estimated value) increase, the above protection function may be turned on to suppress element heat generation, and the power converter 10 (main circuit 50) may be operated to limit the maximum current.

[0219] As a third example, the power converter 10 (main circuit 50) may be controlled using the estimated circuit parameters. Specifically, the control calculations in the control calculation unit 154 can be changed based on the estimated values ​​of the circuit parameters. For example, in the control calculation unit 154, the values ​​of the P (proportional) gain and I (integral) gain in PI control can be changed based on the estimated values ​​of the circuit parameters, according to a pre-prepared correspondence between estimated circuit parameter values ​​and gain values.

[0220] Alternatively, if the control calculation unit 154 is performing a control calculation that directly uses circuit parameters, such as model predictive control, it is possible to update the values ​​of the parameters used in the control calculation in response to changes (increases or decreases) in the estimated values ​​of the circuit parameters. By updating the circuit parameters in this way, the error between the calculation result and the output can be reduced, and the control performance can be improved. In particular, when sensorless control is performed on a motor as the load 5, controllability can be improved by reflecting the estimated value of the inductance L in the control formula. Furthermore, when the load 5 is a power system, controllability can be improved by adjusting the control gain to match the estimated value of the system impedance of the load 5.

[0221] Furthermore, it is possible to create a simulation model of the power converter using the estimated circuit parameters. Analysis using this high-precision simulation model can contribute to the configuration of a digital twin system.

[0222] Embodiment 5. In Embodiments 1 to 4, an example was described in which circuit parameter estimation is performed inside the power converter as part of the functions of the arithmetic unit 100. However, circuit parameter estimation may also be performed outside the power converter.

[0223] Figure 19 is a block diagram illustrating a first arrangement example of the estimation device according to Embodiment 5. As shown in Figure 19, in Embodiment 5, the power conversion device 10Y is connected between the power source 2 and the load 5 to convert the power from the power source 2 into power supplied to the load 5.

[0224] The power converter 10Y is configured from the power converter 10 shown in Figure 2, excluding the estimation unit 200. That is, the power converter 10 can be configured to include the main circuit 50 shown in Figure 2 and a calculation unit 100 capable of realizing the functions of the control unit 150.

[0225] The estimation device 200A is configured to communicate with the arithmetic unit 100 of the power converter 10Y. Communication between the estimation device 200A and the power converter 10Y (arithmetic unit 100) may be performed using either wired or wireless signals. Furthermore, the estimation device 200A is configured to include a data holding unit 210 and an estimation calculation unit 220 similar to the estimation unit 200 in Figure 2. For example, the estimation device 200A can be configured as a computer device. This computer device has a CPU (Central Processing Unit), memory, and a communication interface (not shown), and can be configured to execute the functions of the data holding unit 210 and the estimation calculation unit 220 by software processing through the execution of a pre-stored program.

[0226] From the power converter 10Y, data DAT used for regression calculations for circuit parameter processing, as described in Embodiments 1 to 4, is transmitted to the estimation device 200A. In Embodiments 1 to 4, data DAT includes an output voltage detection value Voref, a current detection value Iref, and a time ratio DTY, which are sampled by the ADC 152 and output to the data holding unit 210. Alternatively, data DAT may further include an input voltage detection value Viref.

[0227] As a result, the estimation device 200A's data holding unit 210 can acquire and hold data for regression calculations similar to those in Embodiment 1. The estimation calculation unit 220 uses the data held in the data holding unit 210 to perform circuit parameter estimation in the same manner as in Embodiments 1 to 4. As a result, the estimation device 200A can output the circuit parameter estimate value PCest, similar to that in Figure 2, to the outside of the estimation device 200A. Furthermore, the above-mentioned abnormality alarm may be output by comparing the circuit parameter estimate value PCest with a predetermined threshold.

[0228] The estimated circuit parameter value PCest and abnormal alarms can also be output to the power converter 10Y. This allows the power converter 10Y to perform circuit operation modifications and control tuning using the estimated circuit parameter values ​​described above. It is also possible to stop the operation of the power converter 10Y (main circuit 50) in response to the abnormal alarm.

[0229] Figure 20 is a block diagram illustrating a second arrangement example of the estimation device according to Embodiment 5. As shown in Figure 20, in Embodiment 5, communication between the power converter 10Y and the estimation device 200A may be performed via a communication network 15 such as the Internet or a dedicated line (not shown). For example, data DAT similar to that shown in Figure 19, output from the power converter 10Y, may be transmitted from the communication unit 11 to the estimation device 200A via the communication network 15.

[0230] In the example shown in Figure 20, the circuit parameter estimate PCest and abnormal alarm obtained by the estimation device 200A can also be transmitted to the power converter 10Y via the communication network 15 and the communication unit 11.

[0231] As shown in the example in Figure 20, the estimation device 200A does not need to be placed in close proximity to the power converter 10Y, and can be placed for remote monitoring or remote control. In Figure 20, the estimation device 200A can also be a computer device that constitutes a server, and can be placed for the purpose of centrally monitoring or centrally controlling multiple power converters 10Y.

[0232] When processing the circuit parameters of multiple power converters 10Y using a single estimation device 200A, the circuit parameters of each power converter 10Y can be estimated by separately preparing regression equations and circuit parameter estimation equations based on state equations corresponding to the circuit topology of the main circuit 50 of each of the multiple power converters 10Y in advance.

[0233] According to the estimation device of Embodiment 5, since the processing for circuit parameter estimation does not need to be performed in the calculation unit 100 of the power converter 10Y, it is expected that the processing load on the calculation unit 100 will be reduced and costs will be lowered.

[0234] In Embodiment 5, when the average value data within one cycle, as described in Embodiment 3, is used in the regression calculation, the averaging process for obtaining the average value data may be performed by either the power converter 10Y or the estimation device 200A. For example, the processing capabilities of the calculation device 100 (power converter 10Y) and the estimation device 200A can be compared, and the averaging process can be performed by the one with higher performance. Alternatively, if the number of data points to obtain the average value data is large (the number of samples within one cycle is large), the averaging process can be performed on the power converter 10Y side to reduce the amount of data transmitted.

[0235] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of this disclosure is indicated by the claims rather than the foregoing description, and all modifications within the meaning and scope of equivalents of the claims are intended.

[0236] 2 Power supply, 5 Load, 10, 10Y Power converter, 11 Communication unit, 15 Communication network, 20 Rectifier circuit, 50, 50A-50F, 50X Main circuit, 51, Q1-Q4 Switching element, 51x Full bridge circuit, 52 Diode, 53 Reactor, 54, 93 Capacitor, 55 Transformer, 61, 62 Boost converter, 71 Current detector, 72, 73 Voltage detector, 80 Neural network model, 90 Average value calculation circuit, 91 Reset switch, 92 Transconductance amplifier, 100 Calculation unit, 150 Control unit, 154 Control calculation unit, 156 Generation unit, 200 Estimation unit, 200A Estimation device, 210 Data holding unit, 220 Estimation calculation unit, C Capacitance value, CW Carrier wave, DAT Data, DTY Ratio, DTY Time ratio, IL, IL1, IL2 Reactor current, Iref Current detection value, Itr Transformer current, L, L1, L2, Ltr Inductance value, NL11~NL1K, NL2 Neuron, PCest Circuit parameter estimate value, Rst Reset signal, Rtr Resistance component (transformer), Spwm PWM signal, Ts Switching period, Vave Average voltage (average value calculation circuit), Vdet Input voltage (average value calculation circuit), Viref Input voltage detection value, Vo Output voltage, Voref Output voltage detection value.

Claims

1. A power conversion device that converts power from a power source into power supplied to a load, comprising: a main circuit connected between the power source and the load and comprising one or more reactors or transformers, one or more switching elements, and one or more capacitors; a current detector for detecting the current flowing through the reactor or transformer; a first voltage detector for detecting at least one of the output voltage from the main circuit to the load or the voltage of the capacitor; a control unit for controlling the on / off state of the switching element to control the output from the main circuit to the load; and an estimation unit for estimating the circuit parameters of the main circuit, wherein the estimation unit acquires data including the current detection value from the current detector, the voltage detection value from the first voltage detector, and the time ratio of the on / off control of the switching element for multiple periods at intervals corresponding to the switching period of the switching element or the calculation period of the control unit, during a data acquisition period including the non-steady operation of the power conversion device, and obtains estimated values ​​of each state coefficient of the state equation of the main circuit by regression calculation using the acquired data, and further calculates estimated values ​​of the circuit parameters using the estimated values ​​of the state coefficients. A power converter in which the circuit parameters include at least one of the following, which are included in any of the state coefficients: the inductance value of the reactor or transformer, the resistance value of the reactor or transformer, the resistance value of the switching element, the capacitance value of the capacitor, and the resistance value of the capacitor.

2. The power conversion device according to claim 1, wherein the main circuit further comprises a diode, any of the state coefficients comprises at least one of the forward voltage and resistance of the diode, and the circuit parameter further comprises at least one of the forward voltage and resistance of the diode.

3. The power conversion device according to claim 1 or 2, wherein the estimation unit includes a data holding unit for holding the current detection value, the voltage detection value, and the time ratio for each of the multiple periods during the data acquisition period, and an estimation calculation unit for calculating estimated values ​​of the circuit parameters, the estimation calculation unit calculates estimated values ​​of each component of a state matrix whose components are each of the state coefficients by performing a regression calculation using a regression equation derived from the state equation and the data held in the data holding unit, and calculates estimated values ​​of the circuit parameters using a circuit parameter estimation formula that pre-determines the relationship between each of the circuit parameters and one or more of the state coefficients and the estimated values ​​of the components.

4. The power converter according to any one of claims 1 to 3, wherein the current detection value, the voltage detection value, and the time ratio used in the regression calculation are the average values ​​over the switching period or the calculation period, respectively.

5. The power converter according to claim 4, wherein the current detection value and the voltage detection value are sampled multiple times in each of the multiple cycles, and the power converter further comprises a neural network model including an input layer that receives an input value including multiple sampled values ​​of the current detection value or multiple sampled values ​​of the voltage detection value, and an output layer that outputs an estimated value of the average value in the switching cycle or the calculation cycle that includes at least a portion of the multiple sampled values.

6. The power conversion device according to any one of claims 1 to 5, wherein the estimation unit is configured to output an estimated value of the circuit parameter, or a signal based on the estimated value of the circuit parameter, to the outside of the estimation unit.

7. The power conversion device according to claim 6, wherein the control unit modifies the operation of the main circuit based on the estimated values ​​of the circuit parameters output from the estimation unit or the signal.

8. The power conversion device according to claim 7, wherein the control unit modifies the operation of the main circuit, including stopping the power conversion operation of the main circuit by fixing the switching element in the off state.

9. The power converter according to claim 7, wherein the control unit modifies the operation of the main circuit, and includes at least one of the following: changing the switching frequency of the switching element, changing the target value of the output voltage, and changing the current limit.

10. The power conversion device according to claim 6, wherein the control unit is configured to update the time ratio at each calculation period by a control calculation based at least on the voltage detected value of the output voltage, and the control unit changes the value of the gain or parameter used in the control calculation in accordance with the change in the estimated value of the circuit parameter.

11. The power conversion device according to any one of claims 1 to 10, further comprising a second voltage detector for detecting the input voltage to the main circuit, wherein the data further includes input voltage detection values ​​by the second voltage detector acquired for multiple cycles for each switching cycle or calculation cycle during the data acquisition period, and the estimation unit calculates estimated values ​​of the circuit parameters using estimated values ​​of each state coefficient in the state equation obtained by the regression calculation using the data including the current detection value, the voltage detection value, the time ratio, and the input voltage detection value.

12. An estimation device for circuit parameters of a power converter that converts power from a power source into power supplied to a load, wherein the power converter comprises: a main circuit connected between the power source and the load and comprising one or more reactors or transformers, one or more switching elements, and one or more capacitors; a current detector for detecting the current flowing through the reactor or transformer; a first voltage detector for detecting at least one of the output voltage from the main circuit to the load or the voltage of the capacitor; and a control unit for controlling the on / off state of the switching element to control the output from the main circuit to the load, wherein the estimation device is configured to perform the following processes: acquiring data including the current detection value by the current detector, the voltage detection value by the first voltage detector, and the time ratio of on / off control of the switching element for multiple periods at intervals corresponding to the switching period of the switching element or the calculation period by the control unit during a data acquisition period including the non-steady operation of the power converter; and calculating estimated values ​​of each state coefficient of the state equation of the main circuit by regression calculation using the acquired data, and further calculating estimated values ​​of the circuit parameters using the estimated values ​​of the state coefficients. An estimation device in which the circuit parameters include at least one of the following, which are included in any of the state coefficients: the inductance value of the reactor or transformer, the resistance value of the reactor or transformer, the resistance value of the switching element, the capacitance value of the capacitor, and the resistance value of the capacitor.

13. The estimation device according to claim 12, wherein the main circuit further comprises a diode, any of the state coefficients comprises at least one of the forward voltage and resistance of the diode, and the circuit parameter further comprises at least one of the forward voltage and resistance of the diode.

14. The estimation device according to claim 12 or 13, wherein the estimation device includes a data holding unit for holding the current detection value, the voltage detection value, and the time ratio for each of the multiple periods during the data acquisition period, and an estimation calculation unit for performing a process to calculate the estimated values ​​of the circuit parameters, the estimation calculation unit calculates the estimated values ​​of each component of a state matrix whose components are each of the state coefficients by performing the regression calculation using a regression equation derived from the state equation and the data held in the data holding unit, and calculates the estimated values ​​of the circuit parameters using a circuit parameter estimation formula that pre-determines the relationship between each of the circuit parameters and one or more of the state coefficients and the estimated values ​​of the components.

15. The estimation device according to any one of claims 12 to 14, wherein the current detection value, the voltage detection value, and the time ratio used in the regression calculation are the average values ​​over the switching period or the calculation period, respectively.

16. The estimation device according to claim 15, wherein the current detection value and the voltage detection value are sampled multiple times in each of the multiple cycles, and the estimation device further comprises a neural network model including an input layer that receives an input value including multiple sampled values ​​of the current detection value or multiple sampled values ​​of the voltage detection value, and an output layer that outputs an estimated value of the average value in the switching cycle or the calculation cycle that includes at least a portion of the multiple sampled values.

17. The estimation device according to any one of claims 12 to 16, wherein the estimation device is configured to output an estimated value of the circuit parameter, or a signal based on the estimated value of the circuit parameter, to the outside of the estimation device.

18. The estimation device according to claim 17, wherein the estimation device outputs an estimated value of the circuit parameter or the signal to the control unit, and the control unit changes the operation of the main circuit based on the estimated value of the circuit parameter or the signal from the estimation device.

19. The estimation device according to claim 18, wherein the control unit modifies the operation of the main circuit, including stopping the power conversion operation of the main circuit by fixing the switching element in the off state.

20. The estimation apparatus according to claim 18, wherein the control unit modifies the operation of the main circuit, and includes at least one of the following: changing the switching frequency of the switching element, changing the target value of the output voltage, and changing the current limit.

21. The control unit is configured to update the time ratio at each calculation period by a control calculation based at least on the voltage detection value of the output voltage, the estimation device outputs an estimated value of the circuit parameter or the signal to the control unit, and the control unit changes the value of the gain or parameter used in the control calculation in accordance with the change in the estimated value of the circuit parameter, according to claim 17.

22. The power converter further comprises a second voltage detector for detecting the input voltage to the main circuit, the data further includes input voltage detection values ​​by the second voltage detector acquired for multiple cycles for each switching cycle or calculation cycle during the data acquisition period, and the estimation device calculates estimated values ​​of the circuit parameters using estimated values ​​of each state coefficient in the state equation obtained by the regression calculation using the data including the current detection value, the voltage detection value, the time ratio, and the input voltage detection value.

Images