JP2005056190A - Control method, controller, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To stably control a control object having the strong non-linearity of input and output relation, without deteriorating control performance. <P>SOLUTION: An input value and an output value for the control object are preliminarily acquired to be accumulated in a database (steps S1-S2). A target series within a predicted section is generated in actual control, based on the output value in the present point, and a plurality of input series up to one portion of divided sections within the predicted section is generated (steps S3-S5). An output series is predicted thereafter based on the database, as to the input series selected from the plurality of input series, and an evaluation value is calculated using an evaluation function (steps S6-S8). The evaluation values are calculated as to all the generated input series to select the lowest evaluation value of input series (steps S9-S11), and the first input value in the selected input series is output to the control object (step S12). <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a control method, a control apparatus, and a program for a control target (for example, an incinerator or the like) whose input / output relationship is strongly nonlinear.

  Conventionally, a control object with a large time constant and a strong nonlinearity of input / output has been linearized by appropriate approximation and controlled by constructing a linear controller. .

  For example, in the conventional modern control theory, PI control (see Patent Document 1), optimal control (see Patent Document 2), etc. are used as linear approximations, and case-based control is performed based on parameters in various control states. Yes.

Japanese Patent Laid-Open No. 5-71718 JP 2002-278603 A

  However, in the conventional method, for example, a process with a strong non-linearity in various input / output relationships such as dead time, such as combustion characteristics, causes a state in which an error from the linear approximation becomes large. Had the problem of significantly degrading control performance.

  That is, when control is performed based on parameters in various control states using PI control or the like, smooth control cannot be performed at the boundary between parameters, and control performance deteriorates. In addition, when the dead time is included, the gain cannot be increased by PI control, and the control performance is deteriorated. Furthermore, for control objects that are generally difficult to formulate, it is possible that control performance will be extremely deteriorated if control is performed within the framework of modern control theory, and appropriate operation input may not be required in theory.

  The present invention has been made in view of the above problems, and is a control method capable of stably performing control on a control target having a strong nonlinearity without degrading control performance. And a control device and a program.

  In order to solve the above problem, a control method according to the present invention is a control method for controlling an output value of a control target to a target value, and an input value and an output value to the control target in an operation state of the control target in advance. And a step of generating a target sequence from the current output value and the target value for a prediction interval from the current time to reaching the target value, and a plurality of input sequences within the prediction interval. Generating an output sequence using the database for the input sequence, obtaining an evaluation value from an evaluation function for bringing the output sequence closer to the target sequence, and the plurality of input sequences And selecting an input sequence that minimizes the evaluation value.

  A control device according to the present invention is a control device that controls an output value of a control target to a target value, and stores in advance a database of input values and output values to the control target in an operation state of the control target. A means for generating a target sequence from the current output value and the target value for a prediction interval from the current time until reaching the target value; a means for generating a plurality of input sequences in the prediction interval; and the input sequence Means for predicting as an output series using the database, means for obtaining an evaluation value from an evaluation function for bringing the output series close to the target series, and the evaluation value is minimized from the plurality of input series Means for selecting such an input sequence.

  The program according to the present invention is a program for controlling an output value of a controlled object to a target value, and previously storing an input value and an output value to the controlled object in an operation state of the controlled object in a database, from the present time Generating a target sequence from a current output value and the target value for a prediction interval until reaching the target value, generating a plurality of input sequences within the prediction interval, and the database for the input sequence Predicting as an output sequence using a step, obtaining an evaluation value from an evaluation function for bringing the output sequence closer to the target sequence, and selecting an input sequence that minimizes the evaluation value from the plurality of input sequences The step is executed by a computer.

  In the control method, the control apparatus, and the program according to the present invention, it is preferable that the input sequence is an input value from the current time point to a predetermined interval in the prediction interval.

  In the control method, the control apparatus, and the program according to the present invention, the evaluation function may be obtained by adding a square of an error between the output series and the target series and a square of the input series. preferable.

  Further, in the control method, the control device, and the program according to the present invention, the control target is an incinerator, and the output value is the top temperature of the incinerator or the steam amount / temperature / pressure for energy recovery. It is preferable that the input value is an operation input of the incinerator.

  The program according to the present invention is a removable recording medium such as a CD-ROM (Compact Disc Read Only Memory), FD (Floppy (registered trademark) Disk), or MO (Magneto-Optic), or a fixed recording medium such as a hard disk. In addition to being able to be recorded and distributed, it can be distributed via a communication network such as the Internet by wired or wireless telecommunication means.

  The control method according to the present invention is a control method for controlling an output value of a controlled object to a target value, and previously storing an input value and an output value for the controlled object in an operation state of the controlled object in a database; Generating a target sequence from a current output value and the target value for a prediction interval from the current time to reaching the target value, generating a plurality of input sequences within the prediction interval, and the input sequence Predicting an output sequence using the database, obtaining an evaluation value from an evaluation function for bringing the output sequence close to the target sequence, and minimizing the evaluation value from the plurality of input sequences Selecting such an input sequence.

  A control device according to the present invention is a control device that controls an output value of a control target to a target value, and stores in advance a database of input values and output values to the control target in an operation state of the control target. A means for generating a target sequence from a current output value and the target value, a means for generating a plurality of input sequences in the prediction section for the prediction interval from the current time until reaching the target value, and the input sequence Means for predicting as an output series using the database, means for obtaining an evaluation value from an evaluation function for bringing the output series close to the target series, and the evaluation value is minimized from the plurality of input series Means for selecting such an input sequence.

  The program according to the present invention is a program for controlling an output value of a controlled object to a target value, and previously storing an input value and an output value to the controlled object in an operation state of the controlled object in a database, from the present time Generating a target sequence from a current output value and the target value for a prediction interval until reaching the target value, generating a plurality of input sequences within the prediction interval, and the database for the input sequence Predicting as an output sequence using a step, obtaining an evaluation value from an evaluation function for bringing the output sequence closer to the target sequence, and selecting an input sequence that minimizes the evaluation value from the plurality of input sequences The step is executed by a computer.

  According to the control method, the control apparatus, and the program according to the present invention, the target sequence and various input sequences can be obtained by using a database that has been measured and stored in advance in relation to the input / output of the control target during actual control. An output sequence is estimated, and an input sequence that minimizes the evaluation function is selected so that the output sequence is as close as possible to the target sequence. Therefore, for a control target whose input / output relationship is strongly nonlinear, an error from linear approximation is reduced, and stable control is always possible without causing deterioration of control performance due to nonlinearity.

  In the control method, the control apparatus, and the program according to the present invention, it is preferable that the input sequence is an input value from the current time point to a predetermined interval in the prediction interval.

  According to this configuration, an input sequence is selected using a branch and bound method as an optimization method. Therefore, the calculation is facilitated and the calculation time can be greatly shortened, and the control performance can be improved without becoming a local optimum answer.

  In the control method, the control apparatus, and the program according to the present invention, the evaluation function may be obtained by adding a square of an error between the output series and the target series and a square of the input series. preferable.

  According to this configuration, the error between the output sequence and the target sequence, which are parameters to be controlled, and the input sequence are used as the evaluation function. Therefore, by selecting an input series having a small evaluation value, the output series can be made as close as possible to the target series.

  Further, in the control method, the control device, and the program according to the present invention, the control target is an incinerator, and the output value is the top temperature of the incinerator or the steam amount / temperature / pressure for energy recovery. It is preferable that the input value is an operation input of the incinerator.

  According to this configuration, the control method, the control device, and the program according to the present invention can be applied to an incinerator whose control target is a combustion process in which the relationship between input and output includes dead time and various nonlinearities. it can.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

  Here, in this embodiment, the branch and bound method is used as the prediction control method. Further, in the present embodiment, an explanation will be given using an incinerator (fluidized bed type waste incinerator) that controls a combustion process whose input / output relationship is strongly associated with various nonlinearities including dead time.

First, a control device according to the present embodiment will be described with reference to FIGS.
FIG. 1 is a schematic configuration diagram of an incinerator to be controlled. FIG. 2 shows a block diagram of the control device according to the present embodiment. FIG. 3 is a diagram illustrating an input value and an output value to be controlled when the input value is randomly changed in advance. FIG. 4 is a diagram showing data accumulated in the database. FIG. 5 is a diagram illustrating a relationship among a target sequence, an output sequence, and an input sequence.

First, an incinerator that is a control target will be described with reference to FIG.
The incinerator 100 is a fluidized bed waste incinerator in which an operation amount (input value) is a waste input amount, a control amount (output value) is a furnace top temperature, and there is a known dead time from the operation amount to the control amount. is there. The manipulated variable (input value) is a step response, and takes a discrete value such as 0, 0.1, 0.2.
In the control device 1, the furnace top temperature, which is the current controlled variable (output value), and the setting contents of the target trajectory to be described later, that is, the target value of the furnace top temperature, which is the controlled variable (output value), and the target from the current time. A prediction interval that reaches a value is input, and a waste input amount that is an operation amount (input value) is output.

Next, the control apparatus according to the present embodiment will be described with reference to FIG.
The control device 1 includes a database storage unit 2, a target sequence generation unit 3, an input sequence generation unit 4, an output sequence prediction unit 5, an evaluation value calculation unit 6, and an input sequence selection unit 7. .

The database storage unit 2 is for storing in advance in the database the input value and the output value in the operation state to be controlled.
In the present embodiment, the database storage unit 2 applies in advance an input value that increases or decreases the amount of waste input, which is the operation amount, to an incinerator 100 that is a control target, and the output value at that time. Is recorded (see FIG. 3). In the present embodiment, the process from the introduction of dust to the temperature measurement at the top of the furnace in the incinerator 100 is a primary system in which the output value is determined by the output value at one point in time and the input value at the present time. That is, in FIG. 3, the current input value is U (k), the output value at one point in time is X (k−1), and X (k) = aX (k−1) + bU (k) ( a and b are predetermined constants).
And about the incinerator 100 which is a control object, the output value X (k) of the next time point with the output value X (k-1) of the immediately preceding divided section and the current input value U (k). ) Are arranged into a set of points (that is, X (k−1), U (k), and X (k) are set as one set) and stored in the database. The format on the database does not need to be determined in particular, and as shown in FIG. 4, for example, as array data in the form of DATA (X (k−1), U (k) = X (k) Also good.
In the database storage unit 2, it is possible to accumulate the database by experimentally collecting data from the incinerator 100 that is the control target in advance, and the operator can input the waste input amount during the actual operation. It is also possible to capture input value and output value data when operating. In order to accumulate more data in the database, it is desirable to capture data of input values and output values when the operator operates the waste input amount even during actual operation. This is because accumulating more data in the database can improve the accuracy of estimation of the next output value in various control states.

The target sequence generation unit 3 is for generating a target sequence from the current output value and target value for the prediction interval from the current time to reaching the target value.
In the present embodiment, a target trajectory (target series) is generated based on the target value and the prediction interval of the furnace top temperature input in advance by the operator. Here, the target trajectory is obtained by, for example, calculating Z (k) by the following equation using a given target value R and the current output value X, and plotting Z plus X. (See FIG. 5).
Z (k) = αZ (k−1) − (1−α) (R−X) (α is a predetermined constant)

The input sequence generation unit 4 is for generating a plurality of input sequences within the prediction interval.
In the present embodiment, only the value obtained by discretizing the operation residue region in which the amount of waste input is considered is set as the operation amount (input value) (here, 16 patterns of 0.1 increments from 0 to 1.6), and The prediction interval is assumed to be 5 sample intervals future k + 5 from the current time k to the divided interval. The input value is a step response. Since the branch and bound method is used as a predictive control method, the input sequence represents a combination of input values, which are discrete values, and divided sections in a tree shape, and an intermediate stage of the tree (that is, one in the prediction section). Part division).

The output sequence predicting unit 5 is for predicting the input sequence generated by the input sequence generating unit 4 as an output sequence using the database accumulated in the database storage unit 2.
In the present embodiment, the prediction of the output series is based on the database accumulated in the database storage unit 2 and the one-time future corresponding to the current output value X (k) and the one-time future input value U (k + 1). Output value X (k + 1), a future output value X (k + 2) corresponding to a future output value X (k + 1), a future time input value U (k + 2), and so on. Processing is performed by recursively reading all output values corresponding to the input values of the input series. Further, when the output value X (k + 1) at the one-time future corresponding to the current output value X (k) and the input value U (k + 1) at the one-time future is not in the database, the data included within a certain distance is stored. Averaging is performed to obtain a future output value X (k + 1) at one time point. Note that the output series to be predicted is up to a part of the divided sections in the prediction section corresponding to the input series.

The evaluation value calculation unit 6 is for calculating an evaluation value based on the evaluation function.
In the present embodiment, as shown in FIG. 5, the evaluation function is the square of the error (shaded portion in the figure) between the output sequence predicted by the output sequence prediction unit 5 and the target sequence generated by the target sequence generation unit 3. And the square of the input sequence generated by the input sequence generation unit 4 are added. Note that the evaluation value is calculated up to a part of the divided sections in the prediction section corresponding to the input series and the output series.

The input sequence selection unit 7 is for selecting an input sequence that minimizes the evaluation value calculated by the evaluation value calculation unit 6 from a plurality of input sequences generated by the input sequence generation unit 4.
In this embodiment, if the evaluation value is lower than the currently temporarily stored evaluation value, the evaluation value is replaced and stored, the input sequence is stored, and all generated by the input sequence generation unit 4 Check if the input series of was checked. If all the input sequences are not checked, an input sequence that is not checked is selected from the input sequences generated by the input sequence generation unit 4, and the output sequence is again selected by the output sequence prediction unit 5. Prediction is performed, and the evaluation value calculation unit 6 calculates the evaluation value. In addition, when the upper and lower limits of the input sequence are set, and the values that the input sequence can take are discrete values, the number of patterns that the input sequence can take in the prediction interval is limited. Therefore, if the number of patterns of the input series is N, the input series that are not checked are sequentially selected from the N patterns. On the other hand, when all the input series are checked, the first value of the stored input series is output to the incinerator 100 that is the control target.

  In the present embodiment, as described above, the first value U (k + 1) of the input sequence output from the input sequence selection unit 7 is used to operate the incinerator 100 that is actually controlled, and then the database. In the storage unit 2, the input value U (k + 1) and the output value X (k + 1) are accumulated in the database, and the next time point is k + 1 as the current time k.

Next, a control method according to the present embodiment will be described with reference to FIG.
FIG. 6 is a flowchart illustrating the processing procedure of the control method according to the present embodiment.

  As shown in FIG. 6, first, in advance, the output value (in this embodiment) of the incinerator that is the control target when the input value (in this embodiment, the amount of waste input) is randomly changed every predetermined time. The furnace top temperature is acquired (step S1). Then, the input value, the output value acquired corresponding to the input value, and the output value of one point in time are stored as a set in the database (step S2).

  Next, the target trajectory (target series) is calculated from the current output value and the target value for the prediction interval from the current time until reaching the predetermined target value at the time of actually controlling the incinerator that is the control target. Generate (step S3).

  Then, an initial value of the evaluation value is set (step S4).

  Next, the prediction section is divided into a plurality of predetermined intervals, and input values are set for each divided section in a part of the divided section (for example, from the current time to the section immediately after the current time or to the several division section). A plurality of input sequences for the prediction interval are generated (step S5). Here, the input sequence is represented by a combination of an input value that is a discrete value and a divided section. Then, one input series is selected (step S6). That is, an input sequence for which the evaluation value is not checked in steps S7 to S9 to be described later is sequentially selected from a plurality of input sequences.

  Then, using the database stored in step S2, an output sequence is predicted for the input sequence selected in step S5 (step S7). Note that the prediction of the output sequence is performed in a range corresponding to the input sequence in a part of the above-described divided section. Here, the prediction of the output sequence is based on the database, the current output value at the present time and the output value at the one time future corresponding to the input value at the future at the one time point, the output value at the future at the first time point, and the input value at the second time point in the future. The processing is performed by recursively reading out all output values in the prediction interval, such as the output value at the time point 2 corresponding to.

  Then, an evaluation value for the input sequence selected in step S6 is calculated from the output sequence predicted in step S7 and the target sequence generated in step S3 using an evaluation function (step S8). Here, the evaluation value is calculated using an evaluation function obtained by adding the square of the error between the target sequence and the output sequence and the square of the input sequence.

  Next, it is determined whether or not the evaluation value calculated in step S8 is lower than the currently stored evaluation value (step S9). If it is low (step S9: YES), the evaluation values are switched and stored, and the input series is stored (step S10). On the other hand, if not low (step S9: NO), the process proceeds to step S11 as it is.

Subsequently, it is confirmed whether or not all input sequences generated in step S5 have been checked (step S11).
If all the input sequences are not checked (step S11: NO), the process returns to step S6, and in steps S6 to S10, an unknown input sequence is optimized by the optimization method so that the evaluation value is minimized. It is determined by a certain branch and bound method.

  On the other hand, if all the input series are checked (step S11: YES), the first value of the stored input series is output to the incinerator that is the control target as the input value of one time in the future ( Step S12). Then, returning to step S1, the input value of the future at the one time point and the output value of the future at the one time point corresponding thereto are fetched (step S1), The database is updated with the current output value as a set (step S2). Then, the process after step S3 is executed with the output value at the time point 1 as the current output value.

  The execution of the control method according to the present embodiment described above can be executed by being read from a recording medium by a CPU (central processing unit) of a computer as a program according to the present embodiment. .

As described above, in the control method, the control apparatus, and the program according to the present embodiment, the actual input / output relationship of the incinerator 100 that is the control target is measured and stored in advance during actual control. (Refer to the database storage unit 2 of the control device 1 in FIG. 1 and steps S1 to S2 in FIG. 6). Estimate the output sequence for the target sequence and various input sequences so that the output sequence is as close as possible to the target sequence. The input sequence that minimizes the evaluation function is selected (the target sequence generation unit 3, the input sequence generation unit 4, the output sequence prediction unit 5, the evaluation value calculation unit 6, the input sequence of the control device 1 in FIG. The selection unit 7 and steps S3 to S11 in FIG. 6).
As a result, with respect to the incinerator 100, which is a control target whose input / output relationship is strongly associated with nonlinearity, by reducing the error from the linear approximation, stable control is always achieved without causing deterioration of control performance due to nonlinearity. It becomes possible.

Further, in the control method, the control apparatus, and the program according to the present embodiment, an input sequence is selected using a branch and bound method as an optimization method. More specifically, an input series is obtained for only some of the divided sections of the prediction section, and an evaluation value is calculated based on the output series and the target series of only the divided sections up to that point. When the evaluation value is larger than the current evaluation value, that is, when the evaluation is bad, even if the input series of the subsequent divided sections is evaluated, it does not become smaller than the current evaluation value. Therefore, the subsequent divided sections are not evaluated and are pruned (the target sequence generation unit 3, the input sequence generation unit 4, the output sequence prediction unit 5, the evaluation value calculation unit 6, and the input of the control device 1 in FIG. Series selection unit 7 and steps S3 to S11 in FIG. 6).
As a result, since it is not necessary to perform calculation for all the divided sections of the prediction section, calculation is facilitated and the calculation time can be greatly reduced, and control performance can be improved without becoming a local optimum answer.

  Note that the branch and bound method is applied as an optimization method, for example, only a value represented by a plurality of patterns obtained by discretizing a possible operation residue region is set as an operation amount (input value), and a prediction interval is set. When the evaluation function is checked for all the divided sections of the prediction section with respect to the combination of the number of patterns of the operation amount and the number of samples of the prediction section, the operation amount pattern It is desirable that the calculation amount of the number of samples in the prediction interval of the number is required, and it is impossible to actually find the solution. Therefore, when the number of combinations is small, it is also possible to obtain input sequences in all the divided sections and calculate the evaluation function by brute force.

Further, the error and the input sequence between the output sequence and the target sequence, which are parameters to be controlled, are used as the evaluation function (see the evaluation value calculation unit 6 of the control device 1 in FIG. 1 and step S8 in FIG. 6).
As a result, by selecting an input series having a small evaluation value, the output series can be made as close as possible to the target series.

Next, simulation results obtained by applying the control method and control apparatus and the program according to the above-described embodiment will be described with reference to FIGS.
FIG. 7 shows a simulation result when the control method, the control apparatus, and the program according to the present embodiment are applied. FIG. 8 shows a simulation result when the PI control which is the conventional method is applied.

As shown in FIG. 7 and FIG. 8, compared with the case where PI control is applied, the control method, the control apparatus, and the program according to the present embodiment cause deterioration in control performance. It can be seen that stable control is possible.
7 and 8, U (k) is the current input value, X (k) is the current output value, and Y (k) is the output value (Y (k + 3) past three time points from the current time. ) = X (k).

  The preferred embodiments of the present invention have been described above. However, the present invention is not limited to the above-described embodiments, and various modifications can be made as long as they are described in the claims.

  In the above-described embodiment, the fluidized-bed type waste incinerator is a control target, but is not limited thereto. For example, control targets include waste treatment plants such as gasification and melting furnaces, sludge and ash melting furnaces, steelmaking plants such as blast furnaces and converters, water treatment plants, etc. May be.

  In the above-described embodiment, the branch and bound method is used as the optimization method, but is not limited thereto. For example, a genetic algorithm or simulated annealing may be used. However, in consideration of optimization using a huge amount of database, it is preferable to use the branch and bound method from the current computer performance.

  In the above-described embodiment, the evaluation function is obtained by adding the square of the error between the output sequence and the target sequence and the square of the input sequence, but is not limited thereto. For example, in addition to that, a parameter to be controlled can also be used.

  In the above-described embodiment, the input value and output value data are stored in the database in advance, but the present invention is not limited to this. For example, another control method and control apparatus, and an initial operation based on a program, during which data of input values and output values are acquired to create a database, and then the control method, control apparatus, and program of the present invention You may switch to However, it is preferable to store data of input values and output values in a database in advance.

It is a schematic block diagram of the incinerator which is a control object. 1 is a block diagram of a control device according to the present embodiment. It is a figure which shows the input value and output value of a control object when an input value is changed at random in advance. It is a figure which shows the data accumulate | stored in the database. It is a figure which shows the relationship between a target series, an output series, and an input series. It is the flowchart explaining the procedure of the process of the control method which concerns on this Embodiment. It is a simulation result at the time of applying the control method and control device concerning this embodiment, and a program. It is a simulation result at the time of applying PI control which is a conventional method.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Control apparatus 2 Database storage part 3 Target series production | generation part 4 Input series production | generation part 5 Output series prediction part 6 Evaluation value calculation part 7 Input series selection part 100 Incinerator (control object)

Claims (12)

A control method for controlling an output value to be controlled to a target value,
Storing the input value and the output value to the control object in the operation state of the control object in advance in a database;
Generating a target sequence from a current output value and the target value for a prediction interval from the current time to reaching the target value;
Generating a plurality of input sequences within the prediction interval;
Predicting the input sequence as an output sequence using the database;
Obtaining an evaluation value from an evaluation function for approximating the output series to the target series;
Selecting an input sequence that minimizes the evaluation value from the plurality of input sequences.   The control method according to claim 1, wherein the input series is an input value from the current time point to a predetermined interval in the prediction interval.   The control method according to claim 1, wherein the evaluation function is obtained by adding a square of an error between the output series and the target series and a square of the input series. The control object is an incinerator,
The output value is the temperature at the top of the incinerator or the amount of steam, temperature and pressure for energy recovery,
The control method according to claim 1, wherein the input value is an operation input of the incinerator. A control device that controls an output value of a control target to a target value,
Means for preliminarily storing an input value and an output value to the control object in an operation state of the control object in a database;
Means for generating a target sequence from a current output value and the target value for a prediction interval from the current time to reaching the target value;
Means for generating a plurality of input sequences within the prediction interval;
Means for predicting the input sequence as an output sequence using the database;
Means for obtaining an evaluation value from an evaluation function for bringing the output series close to the target series;
And a means for selecting an input sequence that minimizes the evaluation value from the plurality of input sequences.   The control apparatus according to claim 5, wherein the input series is an input value from the current time point to a predetermined interval in the prediction interval.   The control device according to claim 5 or 6, wherein the evaluation function is obtained by adding a square of an error between the output series and the target series and a square of the input series. The control object is an incinerator,
The output value is the temperature at the top of the incinerator or the amount of steam, temperature and pressure for energy recovery,
The control device according to claim 5, wherein the input value is an operation input of the incinerator. A program for controlling an output value to be controlled to a target value,
Storing the input value and the output value to the control target in the operation state of the control target in advance in a database;
Generating a target sequence from a current output value and the target value for a prediction interval from the current time to reaching the target value;
Generating a plurality of input sequences within the prediction interval;
Predicting the input sequence as an output sequence using the database;
Obtaining an evaluation value from an evaluation function for bringing the output series closer to the target series;
A program for causing a computer to execute the step of selecting an input sequence that minimizes the evaluation value from the plurality of input sequences.   The program according to claim 9, wherein the input series is an input value from the present time to a predetermined section in the prediction section.   The control program according to claim 9 or 10, wherein the evaluation function is obtained by adding a square of an error between the output series and the target series and a square of the input series. The control object is an incinerator,
The output value is the temperature at the top of the incinerator or the amount of steam, temperature and pressure for energy recovery,
The program according to any one of claims 9 to 11, wherein the input value is an operation input of the incinerator.

Images