JP6933585B2 - Information processing device, information processing method, computer program, control device - Google Patents

Description

The present invention relates to an information processing device.

In each field, a control method using model predictive control (MPC) is used. Model prediction control is a control method that optimizes some control parameters while predicting future responses at each time (finding the optimum solution). For example, in Patent Document 1, regarding the control of an internal combustion engine using model predictive control, the control element is performed within a finite section by iterative calculation using a model of the control element of the internal combustion engine (for example, turbocharger, exhaust gas recirculation). It is described to optimize the operation of. For example, Patent Document 2 describes that a table for performing vehicle control and an approximate expression for calculating control variables of a vehicle are created by using model prediction control.

Japanese Unexamined Patent Publication No. 2011-253536 Japanese Patent No. 5870632

In Patent Document 1, iterative calculation for finding the optimum solution is performed in real time at each time. Since this iterative calculation requires a huge amount of operations, there is a problem that the processing load is high, the processing takes time, and real-time processing may not be possible. In Patent Document 2, since a table or an approximate expression obtained by iterative calculation is created in advance, real-time iterative calculation is not required. However, when targeting a complex system with initial conditions in which many elements (for example, dozens) are related in order to derive the optimum solution of control parameters, for example, control of an internal combustion engine, it is extremely common. A table of dimensions or an approximate expression is required. It is difficult to create a multidimensional table or an approximate expression by exhaustive trials for the initial conditions including such many elements due to the explosive increase in the number of combinations. There is no mention of any issues.

The present invention has been made to solve the above-mentioned problems, and in the control using model prediction control, the processing load is reduced, the processing time is shortened, and the efficiency of initial condition generation is improved. The purpose is.

The present invention has been made to solve at least a part of the above-mentioned problems, and can be realized as the following forms. An information processing device, a model type storage unit that stores in advance a model formula that models a change in the state of a controlled object unit in response to a change in the operation amount of the actuator, and a time-series signal of the operation amount of the actuator. An initial condition storage unit that stores an initial condition including at least a time-series signal of the state of the control target unit, and an actuator of the actuator in the initial condition with respect to the initial condition at each time in the initial condition storage unit. Applying the optimum operation amount storage unit that stores the optimum operation amount, which is the optimum operation amount, and the time-series signal of the operation amount of the actuator created in advance to the model formula in the model formula storage unit. An initial condition generation unit that generates a time-series signal of a predicted value of the state of the control target unit and stores it in the initial condition storage unit as the initial condition, the initial condition in the initial condition storage unit, and the model. The optimum operation amount is obtained by repeating the evaluation while changing the input operation amount for the state estimated using the model expression in the expression storage unit and the objective function using, and the optimum operation amount storage. An information processing apparatus including a prediction processing unit to be stored in a unit and a learning processing unit for obtaining an approximate expression representing the relationship between the initial conditions in the optimum operation amount storage unit and the optimum operation amount. In addition, the present invention can also be realized in the following forms.

(1) According to one embodiment of the present invention, an information processing device is provided. This information processing device includes a model type storage unit, an initial condition storage unit, an optimum manipulated variable storage unit, an initial condition generation unit, a prediction processing unit, and a learning processing unit. The model type storage unit stores in advance a model type that models a change in the state of the controlled object unit of the internal combustion engine according to a change in the operating amount of the actuator of the internal combustion engine. The initial condition storage unit stores at least an initial condition including a time-series signal of the operation amount of the actuator and a time-series signal of the state of the controlled target unit. The optimum operation amount storage unit stores the optimum operation amount, which is the optimum operation amount of the actuator under the initial condition, in association with the initial condition at each time in the initial condition storage unit. The initial condition generation unit applies the time-series signal of the operation amount of the actuator created in advance to the model formula in the model formula storage unit, thereby generating the time-series signal of the predicted value of the state of the control target unit. It is generated and stored in the initial condition storage unit as the initial condition. The prediction processing unit changes the amount of operation to be input with respect to the objective function using the initial condition in the initial condition storage unit and the state estimated using the model formula in the model formula storage unit. While repeating the evaluation, the optimum operation amount is obtained and stored in the optimum operation amount storage unit. The learning processing unit obtains an approximate expression representing the relationship between the initial condition and the optimum manipulated variable in the optimal manipulated variable storage unit.

According to this configuration, the prediction control unit obtains the optimum operation amount of the actuator of the internal combustion engine according to each initial condition in advance by using the model prediction control, and stores it in the optimum operation amount storage unit. Then, the learning processing unit can obtain in advance an approximate expression expressing the relationship between the initial condition in the optimum manipulated variable storage unit and the optimum manipulated variable. Therefore, according to this configuration, the actual control of the internal combustion engine does not require real-time iterative calculation, and corresponds to each element in the initial condition by using the approximate expression obtained by the learning processing unit. The optimum operating amount of the actuator of the internal combustion engine can be quickly obtained according to each actual value (or estimated value), and the processing load can be reduced and the processing time can be shortened. Further, the initial condition generation unit uses a model formula that models a change in the state of the controlled object unit of the internal combustion engine according to a change in the operating amount of the actuator of the internal combustion engine when generating the initial condition. Therefore, when generating the initial condition, the amount of calculation can be reduced and the efficiency of initial condition generation can be improved as compared with the case where the trial in which each element is comprehensively combined is performed.

(2) In the information processing apparatus of the above embodiment, the learning processing unit may obtain the approximate expression by supervised learning of a neural network using the initial conditions and the optimum operation amount as teacher data. Since neural networks can perform complex function approximations, they are suitable for controlling internal combustion engines that can include many elements as initial conditions. According to this configuration, the learning processing unit obtains an approximate expression using such a neural network, so that the accuracy of the approximate expression can be improved.

(3) The information processing apparatus of the above-described embodiment may further include an experiment planning processing unit that generates the time-series signal of the operation amount of the actuator by using the design of experiments method. According to this configuration, the experimental design processing unit generates a time-series signal of the actuator operation amount by using the design of experiments method, so that it is possible to generate a physically reasonable time-series signal as a combination.

(4) In the information processing apparatus of the above embodiment, the initial condition further includes at least one of a disturbance to the internal combustion engine, an output of the controlled object unit, and a target value of the output of the controlled object unit. When the unit is included and the initial condition includes the disturbance, the model formula models the change in the state according to the operation amount and the change in the disturbance, and the initial condition includes the output. In the case, in the model formula, the state and the change in the output according to the change in the operation amount may be modeled. According to this configuration, various factors such as disturbance to the internal combustion engine, output of the control target unit, target value of output of the control target unit, and the like can be considered as initial conditions.

(5) In the information processing apparatus of the above-described embodiment, a linear equation of state and a non-linear equation of state may be used as the model equation. According to this configuration, linear equations of state and nonlinear equations of state can be used as model equations.

(6) In the information processing apparatus of the above-described embodiment, a non-linear equation constructed by using the NARX model may be used as the model formula. According to this configuration, since the nonlinear equation constructed by using the NARX model is used as the model formula, a highly accurate prediction result can be obtained.

(7) In the information processing apparatus of the above embodiment, as the design of experiments, a first method for generating a signal represented by a combination of a step function and a ramp function and a chirp signal whose frequency changes depending on time are used. Either of the second method of generating the represented signal and the second method may be used. According to this configuration, as an experimental planning method, a first method (for example, the APRBS method) for generating a signal represented by a combination of a step function and a ramp function, and a chirp signal whose frequency changes depending on time are used. Either a second method of generating the represented signal (eg, the Sinusoidal Function method) can be used.

(8) According to one embodiment of the present invention, there is provided an information processing method using a model formula that models a change in the state of a controlled object portion of an internal combustion engine in response to a change in the operating amount of an actuator of the internal combustion engine. NS. In this information processing method, a step of generating a time-series signal of a predicted value of a state of the controlled object portion by applying a time-series signal of an operation amount of the actuator created in advance to the model formula, and the above-mentioned A step of storing the time-series signal of the operation amount of the actuator and the generated time-series signal of the state of the controlled object portion as initial conditions, the initial conditions, and the state estimated using the model formula. By repeating the evaluation while changing the input operation amount for the objective function using A step of associating and storing the obtained optimum operation amount and a step of obtaining an approximate expression representing the relationship between the initial condition and the optimum operation amount are provided. According to this method, in the control of the internal combustion engine using the model prediction control, it is possible to reduce the processing load and the processing time, and to improve the efficiency of initial condition generation.

(9) According to one embodiment of the present invention, a computer program using a model formula that models a change in the state of a controlled object portion of an internal combustion engine according to a change in the operating amount of an actuator of the internal combustion engine is provided. .. In this computer program, a function of generating a time-series signal of a predicted value of the state of the controlled object portion by applying a time-series signal of the operation amount of the actuator created in advance to the model formula, and the actuator. A function of storing the time-series signal of the manipulated variable and the generated time-series signal of the state of the controlled target portion as initial conditions, the initial conditions, and the states estimated using the model formula. With respect to the objective function using, the function of obtaining the optimum manipulated variable, which is the optimum manipulated variable of the actuator under the initial condition, by repeating the evaluation while changing the input manipulated variable, and the initial condition. It also has a function of associating and storing the obtained optimum operation amount, and a function of obtaining an approximate expression expressing the relationship between the initial condition and the optimum operation amount. According to this computer program, in the control of the internal combustion engine using the model prediction control, it is possible to reduce the processing load and the processing time, and to improve the efficiency of initial condition generation.

(10) According to one embodiment of the present invention, there is provided a control device for an internal combustion engine including an actuator and a controlled object portion. In the control device of the internal combustion engine, with respect to the initial condition including at least the time-series signal of the operation amount of the actuator and the time-series signal of the state of the controlled target portion, with respect to the initial condition and the initial condition at each time. Information for acquiring a storage unit that stores an approximate expression representing a relationship with the optimum operation amount, which is the optimum operation amount of the actuator, an actual operation amount of the actuator, and a state of the control target unit. Using the acquisition unit, the acquired operation amount and the state, and the approximate expression in the storage unit, the optimum operation amount corresponding to the actual operation amount and the state is obtained, and the optimum operation amount is obtained. A control unit for operating the actuator according to the above is provided. According to this configuration, the control unit uses the manipulated variable and the state (each actual value or estimated value) acquired by the information acquisition unit and the approximate expression in the storage unit to control the internal combustion engine according to each actual value. The optimum operating amount of the actuator can be quickly obtained, and the processing load can be reduced and the processing time can be shortened.

The present invention can be realized in various aspects. For example, an information processing device, an information processing method, an information processing system, and a computer for obtaining an approximate expression for controlling an internal combustion engine using model prediction control. To distribute programs, internal combustion engine control devices using model predictive control, control methods, control systems, computer programs, internal combustion engine control device creation devices, creation methods, creation systems, computer programs, and each of these computer programs. It can be realized in the form of a server device, a non-temporary storage medium that stores the computer program, and the like.

It is a block diagram of the information processing apparatus as one Embodiment of this invention. It is a figure explaining the predictive learning process. It is a flowchart which shows the process procedure in the predictive learning process. It is a flowchart which shows the process procedure in the predictive learning process. It is a figure explaining each step of the predictive learning process. It is a block diagram of the control device of the internal combustion engine as one Embodiment of this invention. It is a flowchart which shows the process procedure in the internal combustion engine control. An example of operation by internal combustion engine control is shown. An example of the calculation time required for internal combustion engine control is shown.

<Information processing device>
FIG. 1 is a block diagram of an information processing device 1 as an embodiment of the present invention. The information processing device 1 is a device that uses Model Predictive Control (MPC) to obtain an approximate expression for use in controlling an internal combustion engine. In this embodiment, the following five elements a1 to a5 are exemplified as elements used for controlling an internal combustion engine. Since the elements a1 to a5 are used as initial conditions for model prediction control, the elements a1 to a5 are also collectively referred to as "initial conditions". The information processing apparatus 1 of the present embodiment obtains an approximate expression expressing the relationship between the initial condition composed of the elements a1 to a5 and the optimum operation amount (optimum operation amount) of the actuator of the internal combustion engine corresponding to the initial condition. It is a device. The approximate expression obtained by the information processing device 1 is mounted on the control device of the internal combustion engine and used for controlling the internal combustion engine. Details will be described later.
(A1) Operation amount of actuator of internal combustion engine u
(A2) Disturbance w with respect to the internal combustion engine
(A3) State x of the controlled object portion of the internal combustion engine
(A4) Output y of the controlled object portion of the internal combustion engine
(A5) Target value r of the output of the controlled object portion of the internal combustion engine

(A1) The operating quantity u is a physical quantity representing the operating status of one or a plurality of actuators that can be operated in the internal combustion engine. For example, the throttle opening degree, the EGR valve opening degree in the exhaust gas recirculation (EGR) system, and the like correspond to the operation amount u. (A2) The disturbance w is one or more physical quantities that affect the output of the internal combustion engine, and the manipulated quantity u is excluded. For example, the engine speed, the outside air temperature, the outside air pressure, and the like correspond to the disturbance w. (A3) The state x is a physical quantity representing the state of one or more controlled target units included in the internal combustion engine. For example, the exhaust temperature, the exhaust flow rate, and the like in the EGR system correspond to the state x.

(A4) The output y is a physical quantity representing the output of one or more controlled target units included in the internal combustion engine. For example, the EGR rate in the EGR system, the supercharging pressure in the turbocharger, and the like correspond to the output y. (A5) The target value r is a target value of the output (that is, element a4) of one or more controlled target units included in the internal combustion engine. It should be noted that each of the above-mentioned five elements a1 to a5 may include a plurality of items. For example, the disturbance w may include three items of engine speed, outside air temperature, and outside air pressure. Further, the items listed in the above-mentioned five elements a1 to a5 are merely examples, and various items can be adopted.

The information processing device 1 includes a storage unit 100, an information processing unit 200, a ROM, a RAM, and a communication unit (not shown), and each unit is connected to each other by a bus (not shown). The storage unit 100 is composed of a hard disk, a flash memory, a memory card, and the like. The storage unit 100 includes a model type storage unit 110, an initial condition storage unit 120, an optimum manipulated variable storage unit 130, and an approximate type storage unit 140.

The model storage unit 110 models changes in the state x of the controlled object unit of the internal combustion engine and the output y in response to changes in the operating amount u of the actuator of the internal combustion engine and the disturbance w with respect to the internal combustion engine. The model formula is stored in advance. The model formula is obtained by an experiment in advance, and includes current and past information of the manipulated variable u, the disturbance w, the state x, and the output y. How much past information is included depends on the number of time samples in the model formula. As the model equation, for example, a linear state equation, a non-linear state equation, and a non-linear equation constructed by using a NARX (Nonlinear Auto-Regressive eXogenous) model can be used.

The model equation of the continuous time system using the linear equation of state can be expressed as follows for each variable of the manipulated variable u, the disturbance w, the state x, and the output y. Note that t is time, and A, B, C, D, E, and F are constant matrices of appropriate size.
-Equation of state: dx / dt = Ax + Bu + Ew ... (1)
・ Output equation: y = Cx + Du + Fw ・ ・ ・ (2)

The model equation of the discrete-time system using the linear equation of state can be expressed as follows, for example, by using each of the above variables and the discrete-time k.
-Equation of state: x [k + 1] = Ax [k] + Bu [k] + Ew [k] ... (3)
-Output equation: y [k] = Cx [k] + Du [k] + Fw [k] ... (4)

The model formula of the continuous time system using the nonlinear state equation can be expressed as follows, for example, using each of the above variables. Note that f and g are nonlinear vector functions that give outputs of the same dimensions as x and y, respectively. Therefore, it can be said that the nonlinear equation of state includes the linear equation of state.
-Equation of state: dx / dt = f (x, u, w) ... (5)
・ Output equation: y = g (x, u, w) ・ ・ ・ (6)

The model equation of the discrete-time system using the non-linear state equation can be expressed as follows, for example, by using each of the above variables and the discrete-time k.
-Equation of state: x [k + 1] = f (x [k], u [k], w [k]) ... (7)
-Output equation: y [k] = g (x [k], u [k], w [k]) ... (8)

The model equation of the nonlinear equation constructed by using the NARX model can be expressed as follows, for example, by using each of the above variables and the discrete time k. Note that g is a nonlinear vector function that gives an output of the same dimension as y, and is different from the above equation 8 in that a plurality of past times are explicitly sampled.
Y [k + 1] = g (y [k], y [k-1], ..., y [k + 1- _ny ], u [k], u [k-1], ..., u [ k + 1-n _u ]) ・ ・ ・ (9)

In this way, linear state equations, non-linear state equations, and NARX models can be used as model equations.

FIG. 2 is a diagram for explaining the predictive learning process. FIG. 2A shows an example of the initial conditions stored in the initial condition storage unit 120. FIG. 2B shows an example of changes in each element in each iterative cycle of the predictive learning process. The initial condition storage unit 120 of FIG. 1 includes an operation amount storage unit 121, a disturbance storage unit 122, a state storage unit 123, an output storage unit 124, and a target value storage unit 125.

When the operation quantity storage unit 121 represents a change in the physical quantity (in other words, a change in the physical quantity over time) of the operation amount u of the actuator of the internal combustion engine for a plurality of times by the step S110 of the prediction learning process described later. The series signal is stored. In FIG. 2A, time-series signals of two items belonging to the manipulated variable u are illustrated as the manipulated variable u1 and the manipulated variable u2. Similarly, the disturbance storage unit 122 stores a time-series signal representing a change in the physical quantity of the disturbance w for a plurality of times with respect to the internal combustion engine by step S110 of the prediction learning process described later. In FIG. 2A, a time-series signal of one item belonging to the disturbance w is illustrated as the disturbance w1.

The state storage unit 123 stores a time-series signal representing a change in the physical quantity of the state x of the controlled target unit of the internal combustion engine for a plurality of times by step S130 of the prediction learning process described later. In FIG. 2A, a time-series signal of one item belonging to the state x is illustrated as the state x1. Similarly, the output storage unit 124 stores a time-series signal representing a change in the physical quantity of the output y of the control target unit of the internal combustion engine for a plurality of times by step S130 of the prediction learning process described later. In FIG. 2A, a time-series signal of one item belonging to the output y is illustrated as the output y1.

In the target value storage unit 125, a time-series signal representing a change in the physical quantity of the target value r of the output of the control target unit of the internal combustion engine for a plurality of times is stored in step S110 of the prediction learning process described later. In FIG. 2A, the example of the time series signal of the target value r is omitted.

The optimum manipulated variable storage unit 130 is set to the initial conditions (manipulated quantity u, disturbance w, state x, output y, target value r) at each time in the initial condition storage unit 120 by step S200 of the predictive learning process described later. On the other hand, the optimum operating amount (optimal operating amount) of the actuator of the internal combustion engine under the initial conditions is stored in association with each other. The approximate expression storage unit 140 stores the approximate expression obtained from the initial conditions and the optimum operation amount in the optimum operation amount storage unit 130 by step S300 of the prediction learning process described later.

The information processing unit 200 controls each unit of the information processing device 1 by expanding and executing a computer program stored in the ROM in the RAM. In addition, the information processing unit 200 functions as an experiment planning processing unit 210, an initial condition generation unit 220, a prediction processing unit 230, and a learning processing unit 240, and cooperates to execute the prediction learning process described later. The experiment planning processing unit 210 uses the design of experiments method to generate time-series signals of the manipulated variable u, the disturbance w, and the target value r among the initial conditions, and stores them in the initial condition storage unit 120. The initial condition generation unit 220 generates a time-series signal of the state x and the output y among the initial conditions, and stores it in the initial condition storage unit 120.

The prediction processing unit 230 optimally corresponds to the initial conditions (operation amount u, disturbance w, state x, output y, target value r) at each time by model prediction control using the initial conditions in the initial condition storage unit 120. The operation amount is obtained and stored in the optimum operation amount storage unit 130. The learning processing unit 240 obtains an approximate expression by supervised learning of a neural network (NN: Neural Network) using the initial condition and the optimum operation amount in the optimum manipulated variable storage unit 130 as supervised learning, and the approximate formula storage unit 140 obtains an approximate expression. Remember.

3 and 4 are flowcharts showing the processing procedure in the predictive learning process. The predictive learning process generates initial conditions (operation amount u, disturbance w, state x, output y, target value r), and at the same time, the generated initial conditions and the optimum operation amount (optimum operation amount) of the actuator of the internal combustion engine. This is a process for finding an approximate expression that expresses the relationship with. The predictive learning process is executed at an arbitrary timing in the information processing device 1.

FIG. 5 is a diagram illustrating each step of the predictive learning process. In FIG. 5, each element of the initial condition stored as a time-series signal in the initial condition storage unit 120 is schematically shown. The vertical axis represents the name of each element of the initial condition, and the horizontal axis represents each time in the time series signal. Normally, the physical quantity of the corresponding element at the corresponding time is displayed in the table, but in FIG. 5, for convenience of explanation, the display of the physical quantity is omitted and the wording for explanation is described. Further, in FIGS. 3 to 5, for convenience of explanation, a plurality of items that can be included in each element of the initial condition are not distinguished. For example, when the operation amount u includes two items, the operation amount u1 and the operation amount u2, the processing for the operation amount u1 and the operation amount u2 may be the same as the processing for the “operation amount u” described below. ..

In step S100, various initial conditions (operation amount u, disturbance w, state x, output y, target value r) are generated. Specifically, in step S110, the experimental design processing unit 210 uses the design of experiments method to generate time-series signals of the manipulated variable u, the disturbance w, and the target value r, and the manipulated variable storage unit 121 and the disturbance storage unit 122. , Each is stored in the target value storage unit 125. The experiment planning processing unit 210 can use any of the following methods b1 and b2 as the experiment planning method, for example.

(B1) First method for generating a signal represented by a combination of a step function and a ramp function: As the first method, for example, an APRBS (Amplitude modulated Pseudo Random Binary Sequences) method can be used. In the APRBS method, signal levels are continuously handled, and continuous signal combinations are generated in a pseudo-random manner using a Latin super-square design or a D-optimal design that efficiently generates various combinations. In the APRBS method, the length of one section of the signal and the signal change speed at the time of section change can be included in the planning target, so in addition to the combination of signal values, the combination of signal change speeds is also diverse. Can be secured. That is, in the APRBS method, it is possible to plan an experiment including transient changes.

(B2) Second method of generating a signal represented by a chirp signal whose frequency changes with time: As a second method, for example, a Sinemodal Excitation method can be used. In the Sinusoidal Excitation method, a combination of continuous signals is generated using a sinusoidal signal whose frequency changes with time. The rate of change in signal level over time is more diverse than the signal generated by the APRBS method.

In step S120, the prediction calculation unit 221 of the initial condition generation unit 220 includes the model formula stored in the model formula storage unit 110, the time series signal of the operation amount u stored in the operation amount storage unit 121, and the disturbance. The time-series signal of the disturbance w stored in the storage unit 122 is read out. In step S130, the prediction calculation unit 221 of the initial condition generation unit 220 applies (applies) the time-series signals of the read manipulated variable u and the disturbance w to the model formula to obtain the predicted values of the state x and the output y. A time-series signal is generated and stored in the state storage unit 123 and the output storage unit 124, respectively.

As described above, in step S100 of the predictive learning process, by using the model formula, it is possible to avoid the generation of the initial conditions that are unlikely to occur physically and to efficiently generate various initial conditions. As described above, the initial condition is composed of five elements of manipulated variable u, disturbance w, state x, output y, and target value r, but a simple method of generating a combination of these five elements is five elements. The upper and lower limits are set for all, and the five elements are comprehensively combined within the range of the upper and lower limits. However, the initial conditions generated by the exhaustive combination include initial conditions that are extremely unlikely to occur physically. This is because the behavior of the state x and the output y is governed by the physical causal relationship depending on the manipulated variable u and the disturbance w. In this regard, in step S100 of the prediction learning process, it is physically impossible as a combination by using the model formula that reproduces this causal relationship not only from the model prediction control after step S200 but also from the stage of generating the initial condition. It is possible to generate a time-series signal with no state x and output y. This makes it possible to efficiently generate only the initial conditions that can actually occur, and many elements (for example, dozens of elements) are used to derive the optimum solution of the control parameters, for example, in the control of an internal combustion engine. Even in a complex system with an initial condition related to), it is possible to generate an initial condition without causing an explosive increase in the number of combinations.

In step S200, the optimum operation amount corresponding to various initial conditions generated in step S100 is obtained. Specifically, in step S210, the prediction calculation unit 231 of the prediction processing unit 230 reads out the model formula stored in the model formula storage unit 110.

In step S220, the prediction calculation unit 231 of the prediction processing unit 230 is required for future prediction by model prediction control, starting from one time in the time series signal of the initial condition stored in the initial condition storage unit 120. Read the time information. For example, in FIG. 5A, when the time t0 is the starting point, the prediction calculation unit 231 provides information on the current and past times required for prediction, specifically, the initial condition of the current time t0. (Disturbance w, state x, output y, target value r) and the initial condition (operation amount u) of the past time t-1 are read out. Here, the reason why the initial condition of the past time t-1 is read only for the manipulated variable u is that the manipulated variable u at the current time t0 is the prediction target in the later step. The prediction calculation unit 231 may also read out the initial conditions in the past several hours in addition to the current initial conditions for the disturbance w and the like. How far back the information is read out depends on the model formula used by the prediction calculation unit 231.

In step S230, the optimum operation amount corresponding to the initial condition read in step S220 is obtained. Specifically, in step S231, the prediction calculation unit 231 of the prediction processing unit 230 determines a time series of the manipulated variable u, the disturbance w, and the target value r within a finite section from the current time to a predetermined future time. This finite interval corresponds to the "predictive horizon" in model predictive control. For example, as shown in FIG. 5B, the prediction calculation unit 231 determines the manipulated variable u, the disturbance w, and the target value r in a finite interval from the current time t0 to the predetermined future time t5. In the example of FIG. 5B, the prediction calculation unit 231 sets a predetermined default value (FIG. 5: D) for the manipulated variable u, and sets the future time t1 to t5 of the disturbance w and the target value r. , The physical quantity of the current time t0 read in step S220 (FIG. 5: present) is set. In determining the default value of the manipulated variable u, the prediction calculation unit 231 may consider the manipulated variable u of the past time t-1 read in step S220.

In step S232, the prediction calculation unit 231 of the prediction processing unit 230 predicts the state x and the output y in the future finite section by using the model formula. Specifically, the prediction calculation unit 231 sets the condition of the current time as the initial condition (operation amount u, disturbance w, state x, output y, target value r) read out in step S220, and sets the model formula read out in step S210. On the other hand, by applying (applying) the time-series signals of the manipulated variable u and the disturbance w in the finite section determined in step S231, the time-series signals of the predicted values of the state x and the output y in the finite section are applied. To generate. For example, as shown in FIG. 5C, the prediction calculation unit 231 predicts the state x and the output y within a finite interval from a predetermined future time t1 to t5 (FIG. 5: prediction).

In step S233, the evaluation unit 232 of the prediction processing unit 230 includes the time-series signals of the manipulated variable u, the disturbance w, and the target value r in the finite section determined in step S231, and the evaluation unit 232 in the finite section predicted in step S232. The time series signal of the state x and the output y is input to the objective function, and the objective function is evaluated. This objective function is a predetermined formula for quantitatively evaluating the control performance, and is stored in the storage unit 100. As the objective function, for example, the sum of squares of the difference between the output (output y) and the target (target value r) within a finite interval can be used.

In step S234, the iterative processing unit 233 of the prediction processing unit 230 determines whether or not the values of the objective functions have converged. When it converges (step S234: YES), the iterative processing unit 233 shifts the processing to step S236.

On the other hand, when it has not converged (step S234: NO), the iterative processing unit 233 modifies the time-series signal of the manipulated variable u in the finite interval so as to improve the objective function value, and shifts the processing to step S232. Let them repeat the prediction and evaluation. For example, as shown in FIG. 5 (D), the iterative processing unit 233 corrects the manipulated variable u in the finite interval from the current time t0 to the predetermined future time t5, and then the prediction calculation unit 231 corrects the manipulated variable u in FIG. 5 (D). As shown in E), the state x and the output y based on the modified manipulated variable u are predicted, and the evaluation unit 232 evaluates the objective function using these. For example, as shown in FIG. 2B, as the manipulated variable u1 and manipulated variable u2 in the finite interval (predicted horizon HP) are modified to repeat cycles C1, C2, C3, the corresponding finite interval (predicted horizon). The state x1 and the output y1 in HP) also change as shown in cycles C1, C2, and C3. In the third cycle C3 in which the values of the objective functions have converged, it can be seen that the time-series signal of the output y1 substantially matches the time-series signal of the target value r. The iterative processing unit 233 may use known methods such as a gradient method, a shooting method, and a C / GMRES method.

In step S236, the iterative processing unit 233 of the prediction processing unit 230 extracts the operation amount u for one hour as the starting point as the “optimal operation amount” from the time series signal of the operation amount u in the finite interval. Then, the iterative processing unit 233 associates this optimum operation amount with the initial conditions (operation amount u, disturbance w, state x, output y, target value r) read in step S220, and the optimum operation amount storage unit. Store in 130. For example, as shown in FIG. 5 (F), the iterative processing unit 233 sets the operation amount u at the current time t0 as the optimum operation amount, and sets the initial condition read in step S220 (the operation amount u at the past time t-1, the present). The disturbance w at time t0, the state x, the output y, and the target value r) are associated with each other and stored in the optimum manipulated variable storage unit 130.

In step S240, the iterative processing unit 233 of the prediction processing unit 230 determines whether or not the processing for all the times has been completed for the time series of the initial conditions stored in the initial condition storage unit 120. When the processing for all the times is not completed (step S240: NO), the iterative processing unit 233 advances the time starting from the starting point by one hour in step S250, and repeats the processing after step S220. When the processing for all the times is completed (step S240: YES), the iterative processing unit 233 shifts the processing to step S300.

As described above, in step S200 of the prediction learning process, model prediction control is executed for various initial conditions generated in step S100, and the optimum operation amount corresponding to the initial conditions at each time is obtained. As described above, the prediction processing unit 230 performs the process of obtaining the optimum operation amount for the initial condition of each time while moving the time that is the starting point for reading the initial condition by one hour at a time for all the time of the time series signal. Do for minutes. Therefore, the prediction processing unit 230 can obtain the optimum operation amount for various initial conditions and store it in the optimum operation amount storage unit 130.

In step S300, an approximate expression expressing the relationship between the initial condition stored in the optimum manipulated variable storage unit 130 and the optimum manipulated variable is obtained by machine learning. Specifically, in step S310, the NN calculation unit 241 of the learning processing unit 240 reads out all the sets of the initial conditions and the optimum operation amount stored in the optimum operation amount storage unit 130. In step S320, the NN calculation unit 241 of the learning processing unit 240 initializes the parameters in the neural network (NN).

In step S330, the NN calculation unit 241 of the learning processing unit 240 obtains an approximate expression by supervised learning of NN using the initial condition and the optimum operation amount as teacher data. Specifically, the NN calculation unit 241 gives (applies) the initial conditions read in step S310 to the NN, and obtains the output of the NN. In step S340, the evaluation unit 242 of the learning processing unit 240 inputs the output of the NN, the initial condition and the optimum operation amount read in step S310 into the objective function, and evaluates the objective function to optimize the NN. Evaluate the approximation accuracy (error) of the manipulated variable. As this objective function, for example, the sum of squares of the difference between the output of NN and the read optimum manipulated variable can be used.

In step S350, the iterative processing unit 243 of the learning processing unit 240 determines whether or not the values of the objective functions have converged. When it converges (step S350: YES), the iterative processing unit 243 shifts the processing to step S370. On the other hand, when it has not converged (step S350: NO), the iterative processing unit 243 modifies the NN parameter so as to improve the objective function value, shifts the processing to step S330, and repeats the NN output and the evaluation. .. The iterative processing unit 243 may use a known method such as backpropagation.

In step S370, the iterative processing unit 243 of the learning processing unit 240 stores the latest NN and the parameters of the NN in the approximate expression storage unit 140 as learned NNs, and ends the processing.

As described above, in step S300 of the predictive learning process, an approximate expression expressing the relationship between the initial condition stored in the optimum manipulated variable storage unit 130 and the optimal manipulated variable is obtained by machine learning. The relationship between the initial condition generated in step S200 and the optimum manipulated variable can be expressed as a non-linear function in which the optimum manipulated variable is determined once the initial condition is determined. In step S300, an approximate expression is created in advance from this relationship. By using this approximate expression, in the actual control of the internal combustion engine, it is possible to obtain the optimum operation amount at high speed in real time at each time without performing iterative calculation (model prediction control) for finding the optimum solution. It becomes. Here, the relationship between the initial condition and the optimum manipulated variable is generally extremely non-linear, and since many elements are related to the initial condition, it is a high-dimensional relationship with multiple inputs and multiple outputs. Good approximation accuracy cannot be expected with general line formats and nth-order polynomials. In this regard, in step S300 of the predictive learning process, machine learning by NN is used as a method capable of efficiently approximating strong non-linearity with multiple inputs and multiple outputs. When learning the relationship between the initial condition and the optimum manipulated value using NN, the effect of the parameters in the NN on the objective function value can be calculated by backpropagation, and based on that information, iterative calculation based on the gradient method is non-linear. Learning that reproduces the relationship well becomes possible. In this way, the NN learned with sufficient accuracy can realize the control performance (prediction performance) equivalent to the real-time model prediction control with a low calculation load.

As described above, according to the information processing apparatus 1, the prediction processing unit 230 uses the model prediction control to set each initial condition (operation amount u, disturbance w, state x, output y, target value r) in advance. The optimum operating amount of the actuator of the internal combustion engine corresponding to the response is obtained and stored in the optimum operating amount storage unit 130 (FIG. 3: step S200). Then, the learning processing unit 240 can obtain in advance an approximate expression (learned NN and parameters) expressing the relationship between the initial condition and the optimum operation amount in the optimum operation amount storage unit 130 (FIG. 4: Step S300). ). Therefore, according to the information processing device 1 of the present configuration, the actual control of the internal combustion engine does not require real-time iterative calculation, and uses the approximate expression (learned NN and parameters) obtained by the learning processing unit 240. By doing so, the optimum operating amount of the actuator of the internal combustion engine can be quickly obtained according to each actual value (or estimated value) corresponding to each element in the initial condition, and the processing load can be reduced and the processing time can be shortened. Can be planned. Further, when the initial condition generation unit 220 generates the initial condition, the state x (and output y) of the controlled target unit of the internal combustion engine according to the change in the operation amount u (and the disturbance w) of the actuator of the internal combustion engine. Use a model formula that models the change in. Therefore, when generating the initial condition, the amount of calculation can be reduced and the efficiency of initial condition generation can be improved as compared with the case where the trial in which each element is comprehensively combined is performed.

<Control device for internal combustion engine>
FIG. 6 is a block diagram of a control device 3 for an internal combustion engine as an embodiment of the present invention. The control device 3 is a device that controls an internal combustion engine by using an approximate expression created by the information processing device 1. Examples of the internal combustion engine include a diesel engine and a gasoline engine. The control device 3 includes a driver interface (IF) 310, an ECU (Electronic Control Unit) 320, and a hardware system 330, and each part is connected to each other by an in-vehicle network (not shown). The driver interface 310 is an interface for acquiring an operation signal by the driver of the internal combustion engine, and is, for example, an accelerator, a brake, or the like.

The ECU 320 is a microcontroller (microcomputer) that controls the operation of the internal combustion engine. In addition, the ECU 320 functions as a target value determination unit 321, a control unit 322, and an information acquisition unit 323, and cooperates to execute internal combustion engine control described later. The target value determining unit 321 determines the target value r of the output of the controlled target unit of the internal combustion engine according to the operation signal by the driver.

The control unit 322 obtains the optimum operation amount of the actuator 331 by using the approximate expression (learned NN and parameters) obtained by the information processing device 1, and operates the actuator 331. The approximate expression may be stored in advance in a storage unit (not shown) in the control unit 322 at the time of manufacturing the vehicle equipped with the control device 3, for example. Further, the control unit 322 may communicate with the information processing device 1 via a communication network (not shown), periodically acquire an approximate expression, and store it in a storage unit in the control unit 322.

The information acquisition unit 323 acquires each actual value of each element (operation amount u, disturbance w, state x, output y) corresponding to the initial condition based on the detection value acquired by the sensor 333. The information acquisition unit 323 obtains each estimated value of the manipulated variable u, the disturbance w, the state x, and the output y from the actual value acquired by the sensor 333 by using an observer or a Kalman filter based on the model to be controlled. You may. In the control of an internal combustion engine, many elements (for example, several tens) are required to obtain the optimum manipulated variable. Therefore, the information acquisition unit 323 may use the actual value acquired by the sensor 333 for some elements and the estimated value estimated for the other elements. The target value r corresponding to the initial condition is separately set by the target value determining unit 321.

The hardware system 330 is hardware mounted on an internal combustion engine. The actuator 331 is one or more actuators that can be operated in an internal combustion engine, such as a throttle, an EGR valve in an EGR system, and the like. The control target unit 332 is one or a plurality of control target units included in the internal combustion engine, such as an EGR system and a supercharger. The sensor 333 is a sensor for detecting each of the operating status (operation amount u) of the actuator 331, the state x of the controlled object unit 332, the output y of the controlled object unit 332, and the disturbance w with respect to the internal combustion engine.

FIG. 7 is a flowchart showing a processing procedure in internal combustion engine control. The internal combustion engine control is a process of controlling the internal combustion engine (specifically, the actuator 331) by using the optimum operation amount obtained by using the approximate expression created by the information processing apparatus 1. The internal combustion engine control shown in FIG. 7 is started, for example, when the vehicle equipped with the control device 3 is started, and is repeatedly executed at predetermined control cycles.

In step S410, the driver interface 310 acquires an operation signal from the driver of the internal combustion engine from the accelerator opening and the brake opening, and transmits the operation signal to the target value determining unit 321. In step S420, the target value determination unit 321 of the ECU 320 uses the acquired operation signal (operation signal by the driver) and the current operation amount u, disturbance w, state x, and output y acquired from the information acquisition unit 323. The target value r with respect to the output y of the control unit 322 is determined and transmitted to the control unit 322.

In step S430, the control unit 322 of the ECU 320 approximates the acquired target value r and the current manipulated variable u, disturbance w, state x, and output y acquired from the information acquisition unit 323 (learned NN and parameter). Apply to. As a result, the control unit 322 can determine the optimum operation amount (optimum operation amount) of the actuator 331 of the internal combustion engine for achieving the target value r specified by the target value determination unit 321. In step S440, the control unit 322 of the ECU 320 operates the actuator 331 with the determined optimum operation amount.

In step S450, the state x and the output y of the control target unit 332 change as a result of the operation of the actuator 331. In step S460, the sensor 333 detects and detects the latest information regarding the operating status (operation amount u) of the actuator 331, the state x of the control target unit 332, the output y of the control target unit 332, and the disturbance w with respect to the internal combustion engine. The signal is transmitted to the information acquisition unit 323. In step S470, the information acquisition unit 323 of the ECU 320 obtains each actual value or each estimated value of the operation amount u, the disturbance w, the state x, and the output y based on the acquired latest information.

FIG. 8 shows an example of operation by internal combustion engine control. In FIG. 8, the operation amount u is 3 items of operation amounts u1 to u3, the disturbance w is 4 items of disturbance w1 to w4, the state x is 7 items of states x1 to x7, and the output y is 2 items of outputs y1 to y2. A specific example is shown in the case where the internal combustion engine control is executed by using the two items of the target values r1 to r2 as the target value r. FIG. 8A illustrates the time-series signals of the manipulated variable u and the disturbance w, and FIG. 8B illustrates the time-series signals of the output y and the target value r. For convenience of illustration, the time series signal of the state x is omitted. As shown in FIG. 8B, the actual value of the output y is the set target value r even in a complicated case of multiple inputs and multiple outputs that requires many elements to obtain the optimum manipulated variable. It can be seen that it follows properly.

FIG. 9 shows an example of the calculation time required for controlling the internal combustion engine. The horizontal axis of FIG. 9 shows the time, and the vertical axis shows the time (ms) required for the calculation by the ECU 320 of the control device 3. As shown in FIG. 9, it can be seen that the calculation time per operation of the ECU 320 is 0.03 ms to 0.06 ms, and is within 0.2 ms at the maximum. For example, in the conventional example in which the internal combustion engine is controlled in real time by using the C / GMRES method known as a high-speed solution of model predictive control, the calculation time is about 60 ms (Hayato Nakata et al., “ Application of C / GMRES model predictive control to diesel engine intake / exhaust systems "). Compared with this conventional example, in the internal combustion engine control of the present embodiment, the calculation speed can be increased by about 1000 times or more.

As described above, according to the control device 3 of the internal combustion engine, the control unit 322 of the ECU 320 includes the actual values or estimated values of the operation amount u, the disturbance w, the state x, and the output y acquired by the information acquisition unit 323. , The optimum operation amount of the actuator 331 of the internal combustion engine according to each actual value or estimated value can be quickly obtained by using the approximate expression in the storage unit of the control unit 322, and the processing load can be reduced and the processing time can be shortened. Can be planned.

<Modified example of this embodiment>
The present invention is not limited to the above-described embodiment, and can be implemented in various aspects without departing from the gist thereof. For example, the following modifications are also possible.

[Modification 1]
In the above embodiment, an example of the configuration of the information processing device is shown. However, the configuration of the information processing device can be modified in various ways. For example, the information processing device may be configured by the cooperation of a plurality of information processing devices arranged on the network. In this case, for example, at least a part of the experiment planning processing unit, the initial condition generation unit, the prediction processing unit, and the learning processing unit may be realized by different information processing devices.

[Modification 2]
In the above embodiment, an example of an element to be considered as an initial condition in model prediction control is given. However, in the model prediction control, at least a part of the elements a1 to a5 to be considered as the initial condition may be omitted, or further other elements may be considered. Specifically, at least a part of the disturbance w of the element a2, the output y of the element a4, and the target value r of the element a5 may be omitted. For example, when the disturbance w is omitted, the parameter of the disturbance w in the model formula can be omitted. Further, the generation of the initial condition of the disturbance w in step S100 of the prediction learning process (FIG. 3) and the consideration of the disturbance w as the initial condition in step S200 can be omitted. Further, consideration of the disturbance w in steps S420, S430, and S470 of the internal combustion engine control (FIG. 7) may be omitted. Similarly, when the output y and the target value r are omitted, consideration for the omitted elements may be omitted for each of the model formula, the predictive learning process, and the internal combustion engine control.

[Modification 3]
In the above embodiment, an example of the predictive learning process is shown (FIGS. 3 and 4). However, the predictive learning process can be modified in various ways. For example, in step S100, a time-series signal such as the manipulated variable u may be generated without using the design of experiments. For example, in step S300, an approximate expression may be obtained by using a means other than NN such as a support vector machine (SVM). For example, steps S100, S200, and S300 are not executed as a series of processes, but may be executed individually.

For example, in the predictive learning process, some of the above-mentioned steps may be omitted, or other steps may be additionally executed. Specifically, for example, in step S100, at least a part of the initial conditions is generated by the first method (design of experiments), and the remaining time series signals are generated by the second method. You may. For example, a plurality of model formulas are stored in advance in the model formula storage unit, and in steps S130 and S232, the model formulas to be applied are changed according to at least some characteristics and actual values of the initial conditions. May be good. For example, after the end of step S300, the generated approximate expression may be delivered to the control device of the internal combustion engine.

[Modification example 4]
In the above embodiment, an example of the configuration of the control device of the internal combustion engine is shown. However, the configuration of the control device of the internal combustion engine can be modified in various ways. For example, the control device may further include the function of the above-mentioned information processing device for obtaining an approximate expression for use in controlling an internal combustion engine by utilizing model prediction control.

[Modification 5]
In the above embodiment, an example of internal combustion engine control is shown (FIG. 7). However, the predictive learning process can be modified in various ways. For example, some of the steps described above may be omitted, or additional steps may be performed. Specifically, for example, after the end of step S470, the acquired actual values of the manipulated variable u, the disturbance w, the state x, and the output y may be transmitted to the information processing apparatus to help improve the accuracy of the approximate expression. ..

Although the present embodiment has been described above based on the embodiments and modifications, the embodiments of the above-described embodiments are for facilitating the understanding of the present embodiment, and do not limit the present embodiment. This aspect may be modified or improved without departing from its spirit and claims, and this aspect includes its equivalents. In addition, if the technical feature is not described as essential in the present specification, it may be deleted as appropriate.

1 ... Information processing device 3 ... Internal engine control device 100 ... Storage unit 110 ... Model storage unit 120 ... Initial condition storage unit 121 ... Operation amount storage unit 122 ... Disturbance storage unit 123 ... State storage unit 124 ... Output storage unit 125 ... Target value storage unit 130 ... Optimal operation amount storage unit 140 ... Approximate type storage unit 200 ... Information processing unit 210 ... Experiment plan processing unit 220 ... Initial condition generation unit 221 ... Prediction calculation unit 230 ... Prediction processing unit 231 ... Prediction calculation unit 232 ... Evaluation unit 233 ... Iterative processing unit 240 ... Learning processing unit 241 ... NN calculation unit 242 ... Evaluation unit 243 ... Iterative processing unit 310 ... Driver interface 321 ... Target value determination unit 322 ... Control unit 323 ... Information acquisition unit 330 ... Hard Wear system 331 ... Actuator 332 ... Control target part 333 ... Sensor

Claims (10)

It is an information processing device
According to the change of the actuator manipulated variable, the model formula storage unit for previously storing a model formula that models the change in the state of the controlled portions,
An initial condition storage unit that stores an initial condition including at least a time-series signal of the operation amount of the actuator and a time-series signal of the state of the control target unit.
An optimum operation amount storage unit that stores the optimum operation amount, which is the optimum operation amount of the actuator in the initial condition, in association with the initial condition at each time in the initial condition storage unit.
By applying the time-series signal of the operation amount of the actuator created in advance to the model formula in the model formula storage unit, a time-series signal of the predicted value of the state of the control target unit is generated, and the initial condition. The initial condition generation unit to be stored in the initial condition storage unit and
Repeating the evaluation of the objective function using the initial condition in the initial condition storage unit and the state estimated using the model formula in the model formula storage unit while changing the input operation amount. A prediction processing unit that obtains the optimum operation amount and stores it in the optimum operation amount storage unit,
A learning processing unit that obtains an approximate expression representing the relationship between the initial condition and the optimum operation amount in the optimum operation amount storage unit, and
Information processing device. The information processing device according to claim 1.
The learning processing unit is an information processing device that obtains the approximate expression by supervised learning of a neural network using the initial conditions and the optimum operation amount as teacher data. The information processing apparatus according to claim 1 or 2, further comprising.
An information processing device including an experiment planning processing unit that generates the time-series signal of the operation amount of the actuator by using the design of experiments method. The information processing apparatus according to any one of claims 1 to 3.
In addition to the initial conditions,
Disturbance, which is a physical quantity that affects the output of the controlled object,
The output of the control target unit and
The target value of the output of the control target unit and
Includes at least some of
When the disturbance is included in the initial conditions, the model formula models the change in the state according to the manipulated variable and the change in the disturbance.
An information processing apparatus in which, when the output is included in the initial conditions, the state and the change in the output are modeled in the model formula according to the change in the manipulated variable. The information processing apparatus according to any one of claims 1 to 4.
An information processing device that uses a linear equation of state and a non-linear equation of state as the model equation. The information processing apparatus according to any one of claims 1 to 5.
An information processing device that uses a nonlinear equation constructed using a NARX model as the model formula. The information processing apparatus according to any one of claims 4 to 6, which is dependent on claim 3 or claim 3.
As the design of experiments, a first method of generating a signal represented by a combination of a step function and a ramp function, and a second method of generating a signal represented by a chirp signal whose frequency changes with time. An information processing device that uses either and. According to the change of the actuator manipulated variable, an information processing method of the change of state of the control object unit utilizing modeled model equation,
A step of generating a time-series signal of a predicted value of the state of the controlled object portion by applying a time-series signal of the operation amount of the actuator created in advance to the model formula.
A step of storing the time-series signal of the operation amount of the actuator and the generated time-series signal of the state of the controlled object portion as initial conditions.
By repeating the evaluation of the objective function using the initial condition and the state estimated by using the model formula while changing the input manipulated variable, the optimum operation of the actuator under the initial condition is performed. The process of finding the optimum amount of operation, which is the amount,
A step of associating the obtained optimum operation amount with the initial condition and storing it.
A step of obtaining an approximate expression expressing the relationship between the initial conditions and the optimum manipulated variable, and
Information processing method. According to the change of the actuator manipulated variable, a computer program using a model model expression changes in the state of the control object unit,
By applying a time-series signal of the operation amount of the actuator created in advance to the model formula, a function of generating a time-series signal of a predicted value of the state of the control target unit is provided.
A function of storing the time-series signal of the operation amount of the actuator and the generated time-series signal of the state of the controlled object portion as initial conditions.
By repeating the evaluation of the objective function using the initial condition and the state estimated by using the model formula while changing the input manipulated variable, the optimum operation of the actuator under the initial condition is performed. The function to find the optimum operation amount, which is the amount, and
A function of associating the obtained optimum operation amount with the initial condition and storing it.
A function to obtain an approximate expression expressing the relationship between the initial condition and the optimum manipulated variable, and
A computer program that includes. An actuator, a control apparatus for engine Ru and a controlled portion,
It was generated by applying the time-series signal of the operation amount of the actuator and the time-series signal of the operation amount of the actuator created in advance to the model formula modeling the change of the state of the control target unit. Optimal operation of the actuator obtained by using the model formula for the initial condition and the initial condition at each time with respect to the initial condition including at least the time-series signal of the predicted value of the state of the controlled object portion. A storage unit that stores an approximate expression that expresses the relationship with the optimum manipulated variable, which is a quantity,
An information acquisition unit that acquires the actual operation amount of the actuator and the state of the control target unit,
Using the acquired operation amount and state and the approximate expression in the storage unit, the optimum operation amount corresponding to the actual operation amount and the state is obtained, and the actuator is set according to the optimum operation amount. The control unit to operate and
Comprising a control device.

Images