Control device, control method, and program

Abstract

To provide a technology capable of achieving control with high stability.SOLUTION: A control device includes a main control part for outputting a next operation amount ua to a control object by model prediction control on the basis of an operation amount and a control amount at each first control period Tc, and a high-speed complementation control part for outputting a next operation amount ub to the control object at each second control period Tf shorter than the first control period Tc, and the high-speed complementation control part calculates a point on an asymptotically stable trajectory with lookahead length Tp as a time constant as an asymptotical target value rf for complementing the control amount at a time t0 and the target value at a time t0+Tp on the basis of the control amount at the time t0 just after the operation amount ua is outputted from the main control part, prescribed lookahead length Tp calculated from a prediction model to be used for the model prediction control, and the target value, and calculates a sum of an average of instantaneous operation change amounts calculated on the basis of a deviation between the asymptotical target value rf and the current control amount and the operation amount ua as the operation amount ub.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to a control device, a control method, and a program.

Background Art

[0002] As one of control methods aimed at making a control amount of a control target follow a target value, model predictive control is known. Further, there is known a technique capable of suppressing a decrease in control performance even when the control period of model predictive control is too long by providing a controller that controls the control target at a control period shorter than the control period of model predictive control (Patent Documents 1 and 2).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the prior art, for example, when the control target changes dynamically, the control may not be stable.

[0005] The present disclosure has been made in view of the above points, and an object thereof is to provide a technique capable of realizing highly stable control.

Means for Solving the Problems

[0006] A control device according to an aspect of the present disclosure is a control device that outputs an operation amount for a control target and makes the control amount of the control target follow a target value, and a first control period T c Every time, based on the operation amount and the control amount, the next operation amount u for the control target is determined by model predictive control aA main control unit that outputs, and a second control period T shorter than the first control period T c For each, a high-speed compensation control unit that outputs the next operation amount u for the control target, and has the high-speed compensation control unit, and the operation amount u from the main control unit f The control amount at the time t0 immediately after being output, a predetermined look-ahead length T calculated from the prediction model used for the model predictive control, and the target value, based on the look-ahead length T b A point on an asymptotically stable trajectory with as a time constant is interpolated between the control amount at the time t0 and the target value at the time t0 + T a To calculate as a provisional target value r, and the average of the instantaneous operation change amount calculated based on the deviation between the provisional target value r and the current control amount and the operation amount u p Calculate as the sum with u as the operation amount u p p f f a b

Effect of the Invention

[0007] A technique capable of realizing highly stable control is provided.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Best Mode for Carrying Out the Invention

[0009] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. In the following, while suppressing a decrease in control performance due to an overly long control period of model predictive control, a control device 10 that can achieve highly stable control even when the controlled object changes dynamically (for example, when the gain of the controlled object changes, etc.) will be described. Note that the control device 10 described below is assumed to be an edge device (e.g., a PLC (Programmable Logic Controller), a DCS (Distributed Control System), etc.) with less computing resources compared to, for example, a PC (Personal Computer).

[0010] Here, generally, when an overly short control period is set for a controlled object with a long time constant, the prediction effect may not be fully exerted. Also, there is a tendency to consume a large amount of memory or fail to correctly identify the model. On the other hand, when an overly long control period is set, the influence of changes in the target value (target value for the control amount) or disturbances that occur during the control timing cannot be reflected in the control, and the control performance may deteriorate. Therefore, although it is necessary to appropriately set a long control period according to the time constant of the controlled object, there is a dilemma that the control performance may deteriorate due to an overly long control period.

[0011] Therefore, in the control device 10 described below, in addition to the controller of model predictive control (main control unit 104 described later), a controller (high-speed complementary control unit 105 described later) that controls the controlled object at a high speed with a control period T c (hereinafter also referred to as "main control period T c ") shorter than the control period T f (hereinafter also referred to as "high-speed control period T f ") is provided. Thereby, even when the main control period T c of model predictive control is long, the high-speed control period T fSince the control target is controlled at high speed every time, for example, it can quickly follow changes in the target value of the control amount or the influence of rapidly changing disturbances, etc., and it is possible to suppress a decrease in control performance. Also, when the computing resources of the control device 10 are scarce and the main control period T c has to be lengthened, it is possible to suppress a decrease in control performance (in other words, even when the computing resources of the control device 10 are scarce, while suppressing a decrease in control performance, a long main control period T c can be used to perform model predictive control).). Furthermore, improvement of control deviation that is insufficiently suppressed by model predictive control (for example, control deviation due to overshoot, etc.) can also be expected. In addition, the stability of the closed loop is improved, and since the consistency between the prediction model of model predictive control and the control target is improved, even when the control target changes dynamically (for example, when the gain of the control target changes, etc.), the robustness (robustness, ruggedness, firmness) against the change can be improved.

[0012] Note that examples of the control amount include the temperature of the control target, and examples of the target value include the set temperature, etc. However, these are all just examples, and the control amount and the target value are not limited to temperature and set temperature, and any control amount in the control target and the target value that is the target of the control amount can be used.

[0013] Hereinafter, assuming a plant as the control target, the case where the control device 10 controls the control target plant 20 will be described. Also, hereinafter, the case where the control target plant 20 changes dynamically according to time (for example, when the gain of the control target plant 20 changes according to time, etc.) is assumed. However, the control target is not limited to a plant, and any device, apparatus, facility, etc. can be used as the control target.

[0014] [First Embodiment] Hereinafter, the first embodiment will be described.

[0015] <Example of the overall configuration of the control device 10> First, an example of the overall configuration of the control device 10 according to the first embodiment will be described with reference to FIG. 1. FIG. 1 is a diagram showing an example of the overall configuration of the control device 10 according to the first embodiment.

[0016] As shown in FIG. 1, the control device 10 according to the first embodiment includes a model parameter estimation unit 101, a measurement unit 102, a differentiator 103, a main control unit 104, a high-speed compensation control unit 105, a first timer 106, a second timer 107, and a switch 108. Each of these units is realized, for example, by a process in which one or more programs installed in the control device 10 are executed by a processor such as a CPU (Central Processing Unit) or an MPU (Micro-Processing Unit).

[0017] The first timer 106 operates as an operation trigger for the model parameter estimation unit 101 and the main control unit 104 every main control period T. c Note that the main control period T is the period for controlling the controlled plant 20 by model predictive control, and its value is set in advance. c

[0018] The second timer 107 operates as an operation trigger for the measurement unit 102 and the high-speed compensation control unit 105 every high-speed control period T. f Note that the high-speed control period T is a period expressed as nT = T, where n is an arbitrary natural number determined in advance, and is the period for controlling the controlled plant 20 by the high-speed compensation control unit 105. The value of the high-speed control period T is also set in advance in the same manner as the main control period T. f f c f c

[0019] The measurement unit 102 is at a high-speed control period T. f ​​​​​​Every time, the control quantity y of the plant 20 to be controlled is measured (observed), and the measured control quantity y is output. Note that the control quantity y of the plant 20 to be controlled is determined according to the manipulated variable u and the disturbance v. Examples of the disturbance v include a decrease or increase in the outside air temperature when the control quantity y is temperature.

[0020] Also, the measurement unit 102 f Every time, acquires (observes) the manipulated variable u output from the switch 108, and outputs the acquired manipulated variable u. Note that examples of the manipulated variable u include the opening / closing amount of a valve that controls the flow rate of a heat medium when the control quantity y is temperature.

[0021] Hereinafter, when explicitly indicating the time t at which the control quantity y is measured, it is expressed as y(t). Similarly, when explicitly indicating the time t at which the manipulated variable u is acquired, it is expressed as u(t). For other variables as well, when explicitly indicating the time t, “(t)” shall be denoted after the variable.

[0022] The differentiator 103 outputs the difference (deviation) between the target value r and the control quantity y as the target deviation e0. The target deviation e0(t) at time t is calculated by e0(t)=r(t)-y(t). Note that hereinafter, as an example, it is assumed that the target value is constant, that is, r(t)=constant. However, this is just an example, and the target value r(t) may vary according to the time t.

[0023] The main control unit 104 c Every time, controls the plant 20 to be controlled by model predictive control. That is, the main control unit 104 c Every time, uses the plant response function {S θ (t)} representing the plant response model of the plant 20 to be controlled to a calculate and output the manipulated variable u for the plant 20 to be controlled. Hereinafter, as an example, a counter called the main control unit update counter is stored in the memory, and every time the manipulated variable u a is calculated and output, the main control unit update counter is updated (for example, incremented by 1). Note that the plant response function {Sθ (t) includes the model parameter θ to be estimated by the model parameter estimator 101.

[0024] Here, the main control unit 104 includes a control parameter calculation unit 111, a look-ahead response correction unit 112, an operation change amount calculation unit 113, and an adder 114.

[0025] The control parameter calculation unit 111 inputs the plant response function {S θ (t)}, and the adjustment coefficients α and β, and calculates the control gain k I and the look-ahead length T p as control parameters, and outputs them to the operation change amount calculation unit 113 and the look-ahead response correction unit 112 respectively. However, the control parameter calculation unit 111 may not be included in the main control unit 104.

[0026] The look-ahead response correction unit 112 corrects the target deviation e0(t) based on the plant response function {S θ (t)}, the target deviation e0(t), the time series data of the change amount du of the past operation amount u a which is the operation change amount time series {du(t)}, and the look-ahead length T p to calculate the corrected target deviation e * (t).

[0027] The operation change amount calculation unit 113 calculates the operation change amount du(t) based on the corrected target deviation e * (t) and the control gain k I . The operation change amount calculation unit 113 calculates and outputs the operation change amount du(t) in the order of, for example, du(t - 3T c ), du(t - 2T c ), du(t - T c ). Note that the operation change amount du is the amount by which the operation amount u c changes every main control period T a .

[0028] The adder 114 adds the operation amount u output from the measurement unit 102 and the operation change amount du acquired from the operation change amount calculation unit 113 to obtain a new operation amount ua Calculate it. That is, for any integer n1, the adder 114 has t = n1T c For each time t satisfying this, a new manipulated variable u a (t)=u(t - T c ) + du(t) is calculated. Then, the adder 114 outputs this new manipulated variable u a to the switch 108.

[0029] The model parameter estimator 101 inputs the control variable y and the manipulated variable u every main control period T c and calculates and outputs an estimated value θ θ of the model parameters of the plant response function {S est (t)}. Hereinafter, the model parameters θ set in the plant response function {S θ (t)} are also referred to as "model parameter set values θ". Note that when calculating the estimated value θ est of the model parameters, an initial value θ0 of the model parameters may be input.

[0030] The high-speed compensation control unit 105 controls the controlled plant 20 every high-speed control period T f . That is, the high-speed compensation control unit 105 calculates and outputs a manipulated variable u f for the controlled plant 20 every high-speed control period T b . However, for times t such that for some integers n2, n3, t = n2T f and t = n3T c , u b (t)=u a (t).

[0031] The switch 108 inputs either the manipulated variable u b or u a and both of u b and outputs the manipulated variable u to the controlled plant 20. That is, the switch 108, taking arbitrary integers as n4, n5, inputs the manipulated variable u c for each time t satisfying t ≠ n4T f and t = n5T b and outputs u = u b , and for t = n4T cand at time t = n5T f For each time t that satisfies a and u b input the operation amount u = u a = u b and output it.

[0032] Note that the control device 10 according to the first embodiment may have a display unit that sequentially displays model parameters and control parameters (control gain, look-ahead length) calculated every main control cycle T c on a display device such as a display, or may have a display control unit that sequentially displays these model parameters and control parameters on the display device of a terminal such as a PC connected by a communication network. As a result, the user or administrator of the control device 10 (hereinafter also referred to as "user, etc.") can know the trends of the model parameters and control parameters, and can confirm the identification status of the plant response model and the adjustment status of the control parameters. Also, it becomes possible to detect a sign when a rapid change occurs in the plant response model.

[0033] <Plant response function {S θ (t)} operation> Next, the operation of the plant response function {S θ (t)} will be described with reference to FIG. 2. FIG. 2 is a diagram for explaining an example of the operation of the plant response function {S θ (t)}.

[0034] As shown in FIG. 2, when the model parameter setting value θ and the time t are input, the plant response function {S θ (t)} outputs the unit step response S θ (t) at time t after the elapse of time t from the initial time 0. Note that the unit step response is the response when the operation amount is a unit step input (that is, the output of the plant response model of the controlled plant 20).

[0035] <Plant response function {S θ (t)} calculation> Next, regarding the process of calculating the unit step response S θ (t) of the plant response function {S θ (t)}, it will be described with reference to FIG. 3. FIG. 3 is a flowchart for explaining an example of the calculation process of the plant response function {S θ (t)}. In FIG. 3, it is assumed that the model parameter setting value θ and the time t are inputted.

[0036] Here, in the present embodiment, regarding the case where an ARMA (autoregressive moving average) model using the autoregression of the past N points for the control amount y and the moving average of the current value and the past M points for the manipulated variable u is adopted as the calculation model of the plant response function S θ (t), it will be described. Note that N and M are, for example, preset by a user or the like. However, this is just an example, and for example, other models such as an ARMAX model may be adopted.

[0037] At this time, let the model parameter setting value θ be

[0038]

Equation

[0039]

Equation

[0040] Step S11: The main control unit 104 initializes the variable representing the time as τ and the state vector φ(k) at the index k as φ(0), and also initializes the index k to k = 0.

[0041] Here, the state vector φ(k) is

[0042] [Number] It is expressed as

[0043] Incidentally, for example, when considering the disturbance v of the past L points, the state vector φ(k) may be expanded to a state vector having L disturbances v(k) as elements in addition to y(k) and u(k).

[0044] The main control unit 104 initializes, for example, τ = 0 and

[0045] [Number] Initializes it as. That is, only the element corresponding to u(k) of φ(0) is initialized to 1, and the other elements are initialized to 0. This means that it is set as the initial value when simulating the unit step response.

[0046] Step S12: Next, the main control unit 104 calculates the control amount prediction value y(k) based on the model parameter setting value θ and the state vector φ(k). The main control unit 104 calculates, for example, y(k)=φ(k) Τ Calculates the control amount prediction value y(k) with θ. Here, Τ represents transpose.

[0047] Step S13: Next, the main control unit 104 updates the state vector φ(k + 1) at the next index (that is, k + 1) based on the control amount prediction value y(k) and the state vector φ(k). The main control unit 104 updates the state vector φ(k + 1), for example, as follows.

[0048] [Number] Here, set the predicted control value calculated in step S12 above for y(k), and set 1 for u(k + 1). Also, set the same values as the state vector φ(k) for y(k - 1), ···, y(k - N + 1) and u(k), ···, u(k - M + 1) (that is, set y(k - 1), ···, y(k - N + 1) and u(k), ···, u(k - M + 1) included in the state vector φ(k) respectively).

[0049] Step S14: Next, the main control unit 104 updates the time τ to τ + ΔT and updates the index k to k + 1. Here, ΔT represents the time width of one step of the ARMA model. The time increment ΔT of the plant response function can be arbitrarily set according to the time constant of the controlled plant 20. For example, it is preferably the same as the main control period T c is preferred.

[0050] Step S15: Next, the main control unit 104 determines whether τ ≥ t. And when it is determined that τ < t (NO in step S15), the main control unit 104 returns to step S12. Thereby, steps S12 to S14 are repeatedly executed until τ ≥ t.

[0051] On the other hand, when it is determined that τ ≥ t (YES in step S15), the main control unit 104 ends the process. Thereby, the finally calculated y(k) is the unit step response S θ (t) is obtained (that is, S θ (t) = y(k) is calculated as the unit step response at time t of the step response function {S θ (t)}).

[0052] <Estimation of model parameter θ> Next, the process of estimating the model parameter θ of the plant response function {S θ (t)} (that is, the model parameter estimated value θ estThe process of calculating (this will be described with reference to FIG. 4). FIG. 4 is a flowchart for explaining an example of the estimation process of the model parameter θ. In FIG. 4, it is assumed that the control amount y(t) and the operation amount u(t) at the current time t are input, and the model parameter estimated value θ est will be calculated for a certain index k. Note that this index k is a value independent of the index k used in the calculation of the plant response function in FIG. 3 and represents the execution-time index of the model parameter estimation process.

[0053] Step S21: The model parameter estimation unit 101 determines whether to initialize the model parameter estimated value θ est (k) and the covariance matrix P(k). Here, cases where it is determined to initialize include, for example, when the model parameter estimated value θ est is calculated for the first time (that is, at the first calculation of the model parameter estimated value θ est ), when an initialization instruction is given by the user or the like, and the like.

[0054] If it is determined to initialize the model parameter estimated value θ est (k) and the covariance matrix P(k) (YES in step S21), the model parameter estimation unit 101 proceeds to step S22. On the other hand, if it is not determined to initialize the model parameter estimated value θ est (k) and the covariance matrix P(k) (NO in step S21), the model parameter estimation unit 101 proceeds to step S23.

[0055] Step S22: If it is determined in step S21 above to initialize the model parameter estimated value θ est (k) and the covariance matrix P(k), the model parameter estimation unit 101 initializes θ est (0) = θ0 and P(0) = I, further initializes the previous control amount value y -1 ← y(t), and initializes k ← 0. Here, as θ0, the plant response function {S θThe model parameter setting value θ set in (t)} may be used, or an initial value assumed in advance by the user or the like may be used, or a fixed initial value such as a vector in which all elements are 0 may be used. Also, as I, it may be an identity matrix or an arbitrary matrix determined in advance.

[0056] Step S23: The model parameter estimation unit 101 updates the state vector φ(k) based on the state vector φ(k - 1), the operation amount u(t), and the previous control amount value y -1 That is, the model parameter estimation unit 101 updates the state vector φ(k) as follows.

[0057]

Equation

[0058] Note that, for example, when considering the disturbance v at the past L points, as described above, the state vector φ(k) is expanded to a state vector having L disturbances v(k) as elements in addition to y(k) and u(k), and the covariance matrix for the disturbance and the estimation of the model parameters may be performed.

[0059] Step S24: Next, the model parameter estimation unit 101 calculates the prediction error ε(k) based on the model parameter estimated value θ est (k - 1), the state vector φ(k), and the control amount y(t). Note that the model parameter estimated value θ est (k - 1) is the estimated value of the model parameter θ estimated in the previous time (that is, at k - 1).

[0060] The model parameter estimation unit 101 calculates the prediction error ε(k) as follows, for example.

[0061] y(k)=y(t) ε(k)=y(k)-φ(k) Τ θ est (k - 1) Step S25: Next, the model parameter estimation unit 101 updates the covariance matrix P(k). The model parameter estimation unit 101 updates the covariance matrix P(k) as follows, for example.

[0062]

Equation

[0063] Step S26: Next, the model parameter estimation unit 101 determines whether to update the estimated value θ est of the model parameters. Here, cases where it is determined to update the estimated value θ est of the model parameters include, for example, cases where the estimated value θ est of the model parameters has not been updated until a predetermined period has elapsed since the initialization in step S22 above, cases where an update instruction is given by the user or the like, and the like.

[0064] If it is determined to update the estimated value θ est of the model parameters (YES in step S26), the model parameter estimation unit 101 proceeds to step S27. On the other hand, if it is determined not to update the estimated value θ est of the model parameters (NO in step S26), the model parameter estimation unit 101 ends without executing step S27.

[0065] Step S27: When it is determined in the above step S26 that the estimated value θ of the model parameters est is to be updated, the model parameter estimation unit 101 updates the estimated value θ est (k) of the model parameters. The model parameter estimation unit 101 updates the estimated value θ est (k) of the model parameters, for example, as follows.

[0066]

Equation

[0067] <Operation of the look-ahead response correction unit 112> Next, the operation of the look-ahead response correction unit 112 will be described with reference to FIG. 5. FIG. 5 is a diagram for explaining an example of the operation of the look-ahead response correction unit 112.

[0068] As shown in FIG. 5, when the plant response function {S θ (t)}, the target deviation e0(t), the time series of operation change amounts {du(t)}, and the look-ahead length T p are input, the look-ahead response correction unit 112 calculates, as the look-ahead response correction value y p (t), the value predicted to change after T n elapses from the current time t due to the past operation change amounts. The look-ahead length T p is calculated by the control parameter calculation unit 111.

[0069] Then, the look-ahead response correction unit 112 outputs the corrected target deviation e n (t) obtained by correcting the target deviation e0(t) with the look-ahead response correction value y * (t). Here, the corrected target deviation e * (t) is calculated as e * (t) = r(t) - (y(t) + y n (t)) = e0(t) - y n (t). The look-ahead response correction value y n(t) can be calculated by the methods described in, for example, International Publication No. 2016 / 092872, Japanese Patent Application Laid-Open No. 2020-21411, etc.

[0070] <Operation change amount calculation unit 113 operation> Next, the operation of the operation change amount calculation unit 113 will be described with reference to FIG. 6. FIG. 6 is a diagram for explaining an example of the operation of the operation change amount calculation unit 113.

[0071] As shown in FIG. 6, the operation change amount calculation unit 113 receives the corrected target deviation e * (t) and the control gain k I When input, the corrected target deviation e * (t) is multiplied by the control gain k I to output the operation change amount du(t). That is, the operation change amount du(t) is calculated as du(t)=k I ×e * (t).

[0072] However, when the result of multiplying the corrected target deviation e * (t) by the control gain k I exceeds the upper limit value du max , the operation change amount calculation unit 113 sets du max as the operation change amount du(t). Similarly, when the result of multiplying the corrected target deviation e * (t) by the control gain k I is below the lower limit value du min , the operation change amount calculation unit 113 sets du min as the operation change amount du(t). Thereby, a limiter for the upper and lower limit ranges can be provided. Note that du a and du max and du min may be set each time so that the operation amount u a after the operation amount u(t) and the operation change amount du are added by the adder 114 does not deviate from the predetermined upper and lower limit ranges.

[0073] <Calculation of control parameters> Next, as control parameters, the control gain k I and the look-ahead length T pThe process of calculating will be described with reference to FIG. 7. FIG. 7 is a flowchart for explaining an example of the calculation process of control parameters. In FIG. 7, it is assumed that the plant response function {S θ (t)} and the adjustment coefficients α and β are input.

[0074] Step S31: The control parameter calculation unit 111 calculates the look-ahead length T θ based on the plant response function {S p and the adjustment coefficient β. The control parameter calculation unit 111 uses the plant response function {S θ (t)} to find the time point when the plant response function value becomes equal to the value obtained by multiplying the plant response function value at the final time T max preset as a sufficiently long value by the adjustment coefficient β as the look-ahead length T θ (T max ). That is, the control parameter calculation unit 111 calculates the look-ahead length T p as follows. p Find T

[0075] , where S p (T θ ) = β × S p (T θ ) max Here, it is preferable that 0 < β ≤ 1. When the response of the closed loop is slow, the speed of the response can be adjusted by setting the adjustment coefficient β to a smaller value.

[0076] Also, the control parameter calculation unit 111 calculates the control gain g p at the time point of the look-ahead length T p as follows.

[0077] g p = S θ (T p ) In this way, the control gain g p is calculated based on the look-ahead length T p . Thereby, for example, even when the dead time and the time constant are long, it becomes possible to respond faster.

[0078] Note that T max may be a value sufficient for the plant response function to converge, or may be a value determined from constraints on memory or the like. Alternatively, the maximum value of the predictable period by the look-ahead response correction unit 112 may be set as T max .

[0079] Step S32: Then, the control parameter calculation unit 111 calculates the control gain k p at the time of the look-ahead length T p based on the control gain g I at that time and the adjustment coefficient α. The control parameter calculation unit 111 calculates the control gain k I as follows.

[0080] [Number] That is, the control gain k p is set to the value obtained by multiplying the adjustment coefficient α by the reciprocal of the control gain g p at the time of the look-ahead length T I . Here, it is preferable that 0 < α ≦ 1. Also, ε is a minute non-negative value for avoiding the control gain k I from becoming too large, and satisfies ε ≧ 0.

[0081] [Calculation outline of the look-ahead length T p Next, the calculation outline of the look-ahead length T p in step S31 above will be described with reference to FIG. 8. FIG. 8 is a diagram for explaining an example of the calculation outline of the look-ahead length T p .

[0082] As shown in FIG. 8, the value of the plant response function S θ (t) gradually changes from time 0 to T max . However, at a certain time T p , S θ (T p ) = β × S θ (T max ) holds. Therefore, such a time T​p is set as the look-ahead length. As a result, since the behavior of the plant at the time of the look-ahead length T p is linked to the final behavior of the plant, it becomes possible to suppress changes in the amount of unnecessary operation, and it is possible to suppress the occurrence of overshoot and undershoot. In addition, it becomes possible to determine a look-ahead length that avoids a period of waste time and the initial period of response where reverse response occurs.

[0083] In the PID (Proportional-Integral-Differential) control known as the prior art, it is necessary to increase the derivative gain according to the magnitude of the dead time, and there is a drawback that it is weak against noise. For example, in the control law by the Ziegler-Nichols step response method, k p = 1.2 / (RL), T I = 2L, T D = 0.5L, R = g p / T p is used, and since the derivative time T D is long and the derivative intensity is large in proportion to the dead time L, the resulting control becomes vulnerable to noise and impulse-like disturbances. In contrast, in this embodiment, since the derivative signal of the observed value is not used, it is possible to realize control that is robust against noise and dead time.

[0084] <Look-ahead length T p Calculation details> Next, the calculation details of the look-ahead length T p in step S31 above will be described with reference to FIG. 9. FIG. 9 is a flowchart for explaining an example of the calculation details of the look-ahead length T p .

[0085] Step S41: First, the control parameter calculation unit 111 sets a variable for exploring the look-ahead length T p as T1 and sets T1 ← 0.

[0086] Step S42: Next, the control parameter calculation unit 111 determines whether T1 ≥ T max is true or not.

[0087] T1 ≥ T max When it is determined that this is the case (YES in step S42), the control parameter calculation unit 111 proceeds to step S46. On the other hand, when it is not determined that T1 ≥ T max is the case (NO in step S42), the control parameter calculation unit 111 proceeds to step S43.

[0088] Step S43: The control parameter calculation unit 111 calculates S θ (T1) ≥ β × S θ (T max ) to determine whether it is satisfied.

[0089] S θ (T1) ≥ β × S θ (T max ) is determined to be satisfied (YES in step S43), the control parameter calculation unit 111 proceeds to step S45. On the other hand, when it is not determined that S θ (T1) ≥ β × S θ (T max ) is satisfied (NO in step S43), the control parameter calculation unit 111 proceeds to step S44.

[0090] Step S44: The control parameter calculation unit 111 updates T1 ← T1 + ΔT and returns to step S42. As a result, T1 ≥ T max or S θ (T1) ≥ β × S θ (T max ) is repeatedly executed until either one is satisfied.

[0091] Step S45: When it is determined that S θ (T1) ≥ β × S θ (T max ) is satisfied, the control parameter calculation unit 111 sets T p ← T1. That is, the control parameter calculation unit 111 sets T1 that satisfies S θ (T1) ≥ β × S θ (T max ) as the look-ahead length T p .

[0092] Step S46: T1 ≥ T max If it is determined that this is the case, the control parameter calculation unit 111 sets T p ← γ × T max That is, where 0 < γ < 1. That is, S θ (T1) ≥ β × S θ (T max ) If T1 that satisfies this cannot be found, the value obtained by multiplying γ by T max is set as the look-ahead length T p .

[0093] <Operation of the High-Speed Completion Control Unit 105> Next, the operation of the high-speed completion control unit 105 will be described with reference to FIG. 10. FIG. 10 is a diagram for explaining an example of the operation of the high-speed completion control unit 105.

[0094] As shown in FIG. 10, for each main control cycle T c , the main control unit 104 outputs the operation amount u a , while the high-speed completion control unit 105 outputs the operation amount u f for each high-speed control cycle T b . However, for a time t such that t = n2T f and t = n3T c for some integers n2 and n3, u b (t) = u a (t). This is because both the main control unit 104 and the high-speed completion control unit 105 are calculated based on the control amount y(t) at the current time t. Hereinafter, the operation amount u a output from the main control unit 104 is also referred to as the "main control operation amount u a ", and the operation amount u b output from the high-speed completion control unit 105 is also referred to as the "complementary operation amount u b ".

[0095] <Calculation of the Complementary Operation Amount u b > Next, the process of calculating the complementary operation amount u b will be described with reference to FIG. 11. FIG. 11 is a diagram for explaining the complementary operation amount u bIt is a flowchart for explaining an example of the calculation process. In FIG. 11, it is assumed that the target value r and the control amount y(t) at the current time t are input, and the complementary operation amount u at a certain index k b is calculated. Note that this index k is an independent value from the index k used for calculating the plant response function in FIG. 3 and the index k used for estimating the model parameter θ in FIG. 4, and represents the execution index of the calculation process of the complementary operation amount u b .

[0096] Step S51: The high-speed complementary control unit 105 determines whether the value of the main control unit update counter has changed (been updated). In other words, the high-speed complementary control unit 105 determines whether the main control operation amount u a is output from the main control unit 104 due to the change in the value of the main control unit update counter. Note that determining whether the main control operation amount u a is output due to the change in the value of the main control unit update counter is just an example, and it may be determined whether the main control operation amount u a is output without using the main control unit update counter. For example, the high-speed complementary control unit 105 may determine whether the main control operation amount u a is output based on whether the main control period T c has elapsed since the previous main control operation amount u a was output.

[0097] If it is determined that the value of the main control unit update counter has changed (YES in step S51), the high-speed complementary control unit 105 proceeds to step S52. On the other hand, if it is not determined that the value of the main control unit update counter has changed (NO in step S51), the high-speed complementary control unit 105 proceeds to step S53.

[0098] Step S52: The high-speed complementary control unit 105 initializes y0 = y(t), m du (0) = 0, and k = 0. Here, m du (k) is the average operation change amount at the index k, as described later.

[0099] Step S53: The high-speed compensation control unit 105 updates the index k as k←k + 1. Since the index k serves as a counter, for example, it may be called a "compensation counter" or the like.

[0100] Step S54: The high-speed compensation control unit 105 calculates a provisional target value r f (k). The provisional target value r f (k) is a provisional target value used by the high-speed compensation control unit 105 and is calculated as follows.

[0101]

Equation

[0102] Step S55: The high-speed compensation control unit 105 calculates an average operation change amount m du (k). The average operation change amount m du (k) is calculated as follows.

[0103]

Equation

[0104] Step S56: The high-speed compensation control unit 105 calculates a compensation operation amount u b . The compensation operation amount u b (k) at the index k is the main control operation amount ua and the average operation change amount m du (k) are used to calculate as follows.

[0105] [Number] Here, u min and u max are the upper and lower constraints imposed on the operation amount. This u min and u max may directly use the upper and lower constraints imposed on the operation amount u a output from the main control unit 104, for example.

[0106] As described above, the complementary operation amount u b (t) = u b (k) at the current time t is obtained.

[0107] <Tentative target value r f (k)> Next, the tentative target value r f (k) calculated in S54 above will be described with reference to FIG. 12. FIG. 12 is a diagram for explaining an example of the tentative target value r f (k).

[0108] As shown in FIG. 12, the tentative target value r f (k) is calculated using the index k as an intermediate point on an asymptotically stable trajectory that reaches the target value r at the look-ahead length T p starting from the control amount y0 = y(t0) at the time t = t0 immediately after the main control unit update counter is updated. That is, the tentative target value r f (k) is calculated as a point on an asymptotically stable trajectory such that the target deviation at the time t0 + T p approaches 0 starting from the control amount y0.

[0109] ><Response when using the high-speed complementary control unit 105> Next, the response when using the high-speed compensation control unit 105 will be described with reference to FIG. 13. FIG. 13 is a diagram for explaining an example of the response when using the high-speed compensation control unit 105.

[0110] As shown in FIG. 13, when the look-ahead length T is controlled by the main control unit 104 such that the target deviation becomes 0 at a previous point in time, even in the response when using the high-speed compensation control unit 105, the look-ahead length T p follows an asymptotically stable trajectory such that the target deviation approaches 0 at a previous point in time. However, the response when using the high-speed compensation control unit 105 does not necessarily coincide with the trajectory using only the main control unit 104, and for example, it may be a trajectory with more suppressed overshoot. p

[0111] <Hardware Configuration of Control Device 10> Next, the hardware configuration of the control device 10 according to the first embodiment will be described with reference to FIG. 14. FIG. 14 is a diagram showing an example of the hardware configuration of the control device 10 according to the first embodiment.

[0112] As shown in FIG. 14, the control device 10 according to the first embodiment includes an input device 201, a display device 202, an external I / F 203, a communication I / F 204, a processor 205, and a memory device 206. These hardware components are communicably connected to each other via a bus 207.

[0113] The input device 201 is, for example, a touch panel or various buttons. The display device 202 is, for example, a display panel. Note that the control device 10 may not have at least one of the input device 201 and the display device 202.

[0114] The external I / F 203 is an interface with an external device such as a recording medium 203a. Examples of the recording medium 203a include an SD memory card (Secure Digital memory card) and a USB (Universal Serial Bus) memory card.

[0115] The communication I / F 204 is an interface for connecting the control device 10 to a communication network. The processor 205 is various arithmetic units such as a CPU or an MPU, for example. The memory device 206 is various storage devices such as an SSD (Solid State Drive), a RAM (Random Access Memory), a ROM (Read Only Memory), or a flash memory, for example.

[0116] Note that the hardware configuration shown in FIG. 14 is an example, and the control device 10 may have other hardware configurations. For example, the control device 10 may have various hardware other than the illustrated hardware.

[0117] [Second Embodiment] Hereinafter, the second embodiment will be described. In the second embodiment, differences from the first embodiment will be described, and descriptions of components that may be the same as those in the first embodiment will be omitted.

[0118] [Example of the overall configuration of the control device 10] An example of the overall configuration of the control device 10 according to the second embodiment will be described with reference to FIG. 15. FIG. 15 is a diagram showing an example of the overall configuration of the control device 10 according to the second embodiment.

[0119] As shown in FIG. 15, the control device 10 according to the second embodiment is obtained by removing the model parameter estimation unit 101 from the overall configuration example described in the first embodiment by setting the pre-estimated model parameter θ in the plant response function {S θ (t)}.

[0120] [Examples] Next, examples will be described.

[0121] In this example, the main control cycle T c = 5 [sec], the high-speed control cycle T fIt was set to 0.2 [sec]. That is, the high-speed compensation control unit 105 was assumed to operate 25 times faster than the main control unit 104. Also, the upper limit u of the operation amount max = 90, and the lower limit u min = 0, and the adjustment coefficients were α = 1.0 and β = 0.5.

[0122] The controlled plant 20 in this embodiment is such that three types of models, model 1, model 4, and model 5, dynamically switch according to time as shown in FIG. 16. That is, model 1 is used from time 0 to 500, model 4 is used from time 500 to 1500, model 1 is used from time 1500 to time 2000, model 5 is used from time 2000 to time 3000, and so on, and the model dynamically switches according to time. Also, it is assumed that model 4 has a gain 0.5 times that of model 1, and model 5 has a gain 2 times that of model 1.

[0123] The step response of model 1 is shown in FIG. 17. As shown in FIG. 17, model 1 takes approximately 150 [sec] until the step response converges. Therefore, for example, when the main control period is T c = 0.5 [sec], approximately 300 prediction buffers are required for predicting until the step response converges. On the other hand, in this embodiment, since the main control period is T c = 5 [sec], only 30 prediction buffers are required for predicting until the step response converges. Therefore, by making the main control period T c = 5 [sec] long, the prediction buffer of the control device 10, which is an edge device or the like, can be saved to about 1 / 10, and both the memory reduction and the calculation time can be shortened.

[0124] The plant response function {S θ (t)} representing the plant response model of the controlled plant 20 in this embodiment is represented by the following equation having an offset term c0.

[0125] y(k)=a1y(k - 1)+a2y(k - 2)+b0u(k)+b1u(k - 1)+b2u(k - 2)+c0 The above equation has elements of two dimensions for the control quantity y and three dimensions for the manipulated variable u. In this embodiment, the disturbance is omitted.

[0126] At this time, the state vector φ(k) is expressed as follows.

[0127]

Equation

[0128]

Equation

[0129] At this time, the control quantity, target value, and manipulated variable in each combination when the high-speed complementary control is enabled or disabled and the model parameter estimation is enabled or disabled will be described. Note that when the high-speed complementary control is enabled, it means the case where the high-speed complementary control unit 105 is used, and when the high-speed complementary control is disabled, it means the case where the high-speed complementary control unit 105 is not used. Also, when the model parameter estimation is enabled, it means the case where the model parameters are estimated online using the model parameter estimation unit 101, and when the model parameter estimation is disabled, it means the case where the plant response model is fixed without using the model parameter estimation unit 101.

[0130] · When the high-speed complementary control is disabled and the model parameter estimation is disabled The control quantity, target value, and manipulated variable in this case are shown in FIG. 18. In FIG. 18, the control quantity (PV) and target value (SP) are shown in the upper row, and the manipulated variable (MV) is shown in the lower row. Also, the right figure is an enlarged view of a part of the time in the left figure. As shown in FIG. 18, in the control using only the main control unit 104, the main control period T cEvery 5 [sec], the manipulated variable changes on the staircase. Also, since the gain of the controlled plant 20 changes to 1 times, 0.5 times, and 2 times according to time, the followability of the controlled variable to the target value has deteriorated.

[0131] ·When the high-speed complementary control is disabled and the model parameter estimation is enabled The controlled variable, target value, and manipulated variable in this case are shown in FIG. 19. In FIG. 19, the controlled variable (PV) and target value (SP) are shown in the upper row, and the manipulated variable (MV) is shown in the lower row. Also, the right figure is an enlarged view of a part of the time of the left figure. As shown in FIG. 19, even when the model parameter estimation is enabled, the control performance has not improved compared to FIG. 18.

[0132] ·When the high-speed complementary control is enabled and the model parameter estimation is disabled The controlled variable, target value, and manipulated variable in this case (that is, the control device 10 according to the second embodiment) are shown in FIG. 20. In FIG. 20, the controlled variable (PV) and target value (SP) are shown in the upper row, and the manipulated variable (MV) is shown in the lower row. Also, the right figure is an enlarged view of a part of the time of the left figure. As shown in FIG. 20, the control performance has been significantly improved compared to FIG. 18, and the followability can be ensured even when the gain of the controlled plant 20 changes to 1 times, 0.5 times, and 2 times according to time, and the robustness against gain fluctuations has been improved.

[0133] Also, as shown in FIG. 20, the manipulated variable is the main control period T c =5 [sec] in addition to the change in the manipulated variable at, and it can be confirmed that a change occurs at the high-speed control period T f =0.2 [sec] due to the high-speed complementary control.

[0134] Therefore, it has been shown that by enabling the high-speed complementary control, it is possible to improve the control performance and the robustness against gain fluctuations of the controlled plant 20 compared to the case where only the main control unit 104 is used.

[0135] ·When high-speed compensation control is enabled and model parameter estimation is enabled The control quantity, target value, and manipulated variable in this case (that is, the control device 10 according to the first embodiment) are shown in FIG. 21. In FIG. 21, the control quantity (PV) and target value (SP) are shown in the upper row, and the manipulated variable (MV) is shown in the lower row. Also, the right figure is an enlarged view of a part of the time in the left figure. As shown in FIG. 21, compared with FIG. 18, the control performance has been greatly improved, and even when the gain of the controlled plant 20 changes to 1 times, 0.5 times, or 2 times according to the time, the followability can be ensured, and the robustness against gain fluctuations has been improved.

[0136] Also, as shown in FIG. 21, it can be seen that the overshoot of the control quantity is slightly reduced compared with FIG. 20. Therefore, by enabling high-speed compensation control, even when model parameter estimation is effective, it is shown that an improvement in control performance and an improvement in the robustness against gain fluctuations of the controlled plant 20 can be achieved compared with the case of using only the main control unit 104.

[0137] ·Changes in each parameter Changes in the model parameters and control parameters when high-speed compensation control is enabled and model parameter estimation is disabled, and changes in the model parameters and control parameters when high-speed compensation control is enabled and model parameter estimation is enabled are shown in FIG. 22. As shown in FIG. 22, when model parameter estimation is effective, at model 4 (gain 0.5 times), there is a behavior of trying to increase the control gain k I . That is, the control gain k I is adjusted up and down according to the increase and decrease of the gain of the controlled plant 20 to achieve a balance. Therefore, it can be seen that even when high-speed compensation control is effective, the online estimation of the model parameters by model parameter estimation functions effectively.

[0138] In this way, when model parameters are estimated by the model parameter estimation unit 101 of the control device 10 according to the first embodiment, the high-speed complementary control unit 105 controls the controlled plant 20 so that the plant reaches an asymptotically stable trajectory having a time constant of a plant response model having the model parameters. Therefore, it is possible to employ a configuration in which model parameter estimation is simultaneously enabled even when high-speed complementary control is enabled.

[0139] [summary] As described above, the control device 10 according to the first and second embodiments has a shorter control period T f The high-speed complementary control unit 105 controls the controlled plant 20 at high speed. c Even if the control period T f Since the controlled plant 20 is controlled at high speed every main control period T, for example, it is possible to quickly follow changes in the target value for the controlled variable and the influence of fast-changing disturbances, and it is possible to suppress deterioration of the control performance. c Even when it is necessary to lengthen the time, it is possible to suppress the deterioration of the control performance. Furthermore, it is expected that the control deviation (e.g., the control deviation due to overshoot, etc.) that is insufficiently suppressed by the model predictive control can be improved. In addition, it is possible to realize highly stable control even when the controlled plant 20 changes dynamically.

[0140] The operation of the main control unit 104 is not limited to that described in the first embodiment, and any operation can be adopted as long as the control method adopts model predictive control. p The asymptotically stable trajectory that eventually reaches the target value r is called the provisional target value r. f As the control gain k I Therefore, it is preferable that main control unit 104 also operates in a similar manner.

[0141] The present invention is not limited to the specifically disclosed above embodiments, and various modifications, changes, combinations with known technologies, etc. are possible without departing from the description of the claims.

Explanation of Signs

[0142] 10 Control device 20 Controlled plant 101 Model parameter estimator 102 Measuring unit 103 Differentiator 104 Main control unit 105 High-speed complementary control unit 106 First timer 107 Second timer 108 Switch 111 Control parameter calculation unit 112 Look-ahead response correction unit 113 Manipulation change amount calculation unit 114 Adder 201 Input device 202 Display device 203 External I / F 203a Recording medium 204 Communication I / F 205 Processor 206 Memory device 207 Bus

Claims

1. A control device that outputs an operation amount for a control target and causes the control amount of the control target to follow a target value, First control period T ｃ At every time, based on the operation amount and the control amount, a main control unit outputs the next operation amount u for the controlled object by model predictive control ａ and The first control period T ｃ A second control period T shorter than ｆ Each time, the next operation amount u for the control target ｂ And a high-speed compensation control unit that outputs, The high-speed compensation control unit, The operation amount u from the main control unit ａ at the time t immediately after being output ０ of the control amount, a predetermined look-ahead length T calculated from the prediction model used for the model predictive control ｐ and based on the target value, a point on an asymptotically stable trajectory having the look-ahead length T ｐ as the time constant, the control amount at the time t ０ and the time t ０ +T ｐ interpolate the target value at to obtain a provisional target value r ｆ and calculate it as the provisional target value r ｆ and the average of the instantaneous operation change amounts calculated based on the deviation between the current control amount and the operation amount u ａ and calculate the sum as the operation amount u ｂ A control device that calculates as

2. The high-speed compensation control unit, the second control period T ｆ Using k as the index updated every time, the control amount y ０ at the time t ０ and the target value r ０ at the time t ｐ + T ｆ r(k) = a(k) × r + (1 - a(k)) × y ０ (where a(k) = 1 - exp( - k × T ｆ / T ｐ )) to calculate the provisional target value r ｆ The control device according to claim 1.

3. The high-speed compensation control unit, the provisional target value r ｆ and a deviation from the control amount at the time t ０ is multiplied by a predetermined gain k Ｉ to calculate a value as the instantaneous operation change amount, the control device according to claim 1

4. The high-speed compensation control unit, the initial value at the time t ０ The control device according to claim 2, wherein an average of the instantaneous operation change amounts with respect to the index k is calculated with the initial value at ０ being set to 0.

5. The gain k Ｉ is calculated based on the prediction model used for the model predictive control, and is the control device according to claim 3.

6. Based on the controlled variable and the manipulated variable observed from the controlled object, the parameters of the prediction model are sequentially estimated every first control period T ｃ The apparatus further includes a model parameter estimation unit that sequentially estimates the parameters of the prediction model every time period T The look-ahead length T ｐ is sequentially updated based on the prediction model having the parameters estimated by the model parameter estimation unit, the control device according to any one of claims 1 to 5.

7. The high-speed compensation control unit, The operation amount u output from the main control unit ａ Using the upper and lower limit ranges applied to the operation amount u that satisfies the upper and lower limit ranges ｂ The control device according to claim 1, which calculates the operation amount u

8. A control device that outputs an operation amount for a control target and causes the control amount of the control target to follow a target value, First control period T ｃ At every time, based on the operation amount and the control amount, a next operation amount u for the control target is output by model predictive control ａ main control procedure, and The first control period T ｃ A second control period T shorter than the first control period T ｆ Each time, the next operation amount u for the control target ｂ Execute a high-speed compensation control procedure for outputting, The high-speed compensation control procedure, At the time t immediately after the operation amount u is output in the main control procedure ａ , based on the control amount at the time t ０ , a predetermined look-ahead length T calculated from the prediction model used for the model predictive control ｐ , and the target value, a point on an asymptotically stable trajectory with the look-ahead length T as the time constant is ｐ calculated as a provisional target value r that interpolates between the control amount at the time t ０ and the target value at the time t ０ +T ｐ , and ｆ is calculated as The provisional target value r ｆ The average of the instantaneous operation change amount calculated based on the deviation between the current control amount and the operation amount u ａ And calculate the sum of the operation amount u ｂ As the control method for calculating as the operation amount u

9. In a control device that outputs an operation amount for a control target and causes the control amount of the control target to follow a target value, First control period T ｃ At every time, based on the operation amount and the control amount, the next operation amount u for the controlled object is output by model predictive control ａ main control procedure, and the first control period T ｃ a second control period T shorter than the first control period T ｆ For each time, a high-speed compensation control procedure for outputting the next operation amount u for the control target ｂ is executed The high-speed compensation control procedure, At the time t immediately after the operation amount u is output in the main control procedure ａ , based on the control amount at that time, a predetermined look-ahead length T calculated from the prediction model used for the model predictive control ０ , and the target value, a point on an asymptotically stable trajectory with the look-ahead length T as the time constant is interpolated between the control amount at the time t ｐ and the target value at the time t + T ｐ to calculate a provisional target value r ０ , and ０ ｐ ｆ calculate it as​​ the provisional target value r ｆ and the average of the instantaneous operation change amounts calculated based on the deviation between the current control amount and the operation amount u ａ as the operation amount u ｂ A program that calculates as

Images