JP2005078545A - Method and device for adjusting process model - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and a device for adjusting a process model which can support a process operation prediction, a process diagnosis, an optimum operation or the like. <P>SOLUTION: In the method for adjusting the process model simulating an actual process, static parameters and dynamic parameters of the process model are adjusted independently of each other on the basis of data obtained from the actual process. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention targets manufacturing processes such as chemical, petrochemical, petroleum, electric power, gas, steel, chemicals, food, waste incineration, recycling, water and sewage, microprocess, etc., process analysis, process operation prediction, process diagnosis A process model adjustment method for adjusting process model parameters online and offline using a process model in order to automatically generate, optimize and optimize the process model and state (operation, optimization, training, etc.) The present invention relates to an adjusting device.

(Conventional example 1)
FIG. 5 is a diagram showing a configuration example of a model construction system used in conventional model prediction control and the like.
In FIG. 5, the operation amount u input to the actual process 1 and the control amount y output from the actual process 1 are taken into the analysis unit 3 via the process interface 2. Further, the process operation data 4 is also taken into the analysis means 3 via the process interface 2. The analysis means 3 analyzes the captured data.

The adjusting unit 5 estimates model parameters based on the linear model 6 based on the analysis result of the analyzing unit 3. In the estimation, for example, the process response is approximated by a first-order lag element and a dead time element. The estimated model parameters are stored in the database 7.
In this way, model parameters are estimated, and process operation prediction, process diagnosis, optimum operation, and the like are performed using the process model in which the estimated model parameters are set.

(Conventional example 2)
In another process model adjustment apparatus, a nonlinear simple model is expressed based on the nonlinearity of the model, and model parameters are estimated using the nonlinear simple model. The non-linear simple model includes an empirical non-linear model, a model constructed by a neural network, fuzzy, or the like.

(Conventional example 3)
In another process model adjustment apparatus, a model based on the mass balance and heat balance of an actual process is constructed, and the parameters of the model are estimated using the model.

JP 2001-249705 A JP 2002-373002 A

  In the conventional example 1, when the operation is performed with a non-linear process, a deviation occurs between the actual process and the process model, which cannot be used for process diagnosis, optimization, and optimum operation.

  In Conventional Example 2, if a small-scale and accurate model is constructed, state estimation and optimization can be performed within a limited range. However, it cannot be used for operation prediction, process diagnosis, or optimum operation at medium scale or higher.

  Conventional Example 3 can be used for many purposes if a physical property calculation method or the like is incorporated, but there are few that can be operated in real time by incorporating a strict actuator, a device model, and a sensor model. Even if it exists, it is only possible to calculate the static parameters of the device (the amount of material and coefficient used to determine the state in which the device is in steady operation) from the steady state data of the process.

  For this reason, the conventional technology has a model that can achieve highly accurate operation prediction, steady state optimization, and optimal operation (such as dynamic optimization) and a model parameter estimation or update method that enables it. There is no system.

  The present invention has been made to solve the above-described problems, has a precise process model, and implements the process model by separately adjusting the static parameters and the dynamic parameters of the process model. It is an object of the present invention to realize a process model adjustment method and adjustment apparatus that can be adjusted so as to be a highly accurate process model close to a process and can support process operation prediction, process diagnosis, optimum operation, and the like.

  In order to achieve such a subject, the present invention is configured as follows.

(1) In a process model adjustment method for adjusting a process model for simulating an actual process,
A method of adjusting a process model, characterized by separately adjusting static parameters and dynamic parameters of a process model based on data obtained from an actual process.

(2) The process model adjustment method according to (1), wherein in the static parameter adjustment, the static parameter is adjusted in consideration of a steady balance of the process model.

(3) The method for adjusting a process model according to (1), wherein in the adjustment of the dynamic parameter, the dynamic parameter is adjusted in advance taking into account the dynamics of how the process moves.

(4) In the adjustment of the dynamic parameter, the dynamic parameter is adjusted in consideration of a response characteristic when a signal is input to the process.

(5) The process model adjustment method according to any one of (1) to (4), wherein the process model has a function of distinguishing a dynamic parameter from a static parameter.

(6) In a process model adjustment device for adjusting a process model for simulating an actual process,
A first adjustment means for adjusting static parameters of the process model based on data obtained from an actual process;
A second adjusting means for adjusting a dynamic parameter of the process model based on data obtained from an actual process;
An apparatus for adjusting a process model, comprising:

As is apparent from the above description, the present invention has the following effects.
It has a precise process model and adjusts static parameters and dynamic parameters of the process model separately. As a result, the process model can be adjusted to be a high-precision process model close to an actual process, and the following effects can be obtained.
(A) A process operation can be predicted with high accuracy.
(B) Continuous process diagnosis can be performed by adjusting conditions and parameters leading to performance degradation and abnormality of the device, and efficient periodic repair, replacement, and safe driving can be performed.
(C) The process model and state automatic generation function can reduce the engineering cost that can be used for many purposes.
(D) It can support optimization of a large-scale plant with high accuracy.
(E) It can support the achievement of advanced and optimal plant operation.

  As described above, according to the present invention, a process model adjustment method and adjustment apparatus capable of supporting process operation prediction, process diagnosis, optimum operation, and the like have been realized.

Hereinafter, the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram showing an embodiment of the present invention.
In FIG. 1, a process model 10 using a precise model is intended for a device having a physical model or a heat quantity model for each device. There is a model that can select physical property information and calculation method of the material to be handled. The process model 10 is a model built according to an actual object based on a simulator, and is realized by what can express the behavior of the plant with high accuracy. In the process model 10, physical parameters and characteristic parameters (dynamic parameters, static parameters) of the device are set, and advanced models are set up to handle material data and physical property calculation methods.

The actual process 11 is considered for the same range as the process model 10. Disturbance information, manipulated variables, controlled variables (including measured values obtained directly, analyzed values obtained by calculating measured values), and computational quantities such as soft sensors and observers can also be used as data in the process model 10 in the actual process 11 . The more data, the more accurate model adjustment is possible.
In the case of off-line, it is sufficient that the data of the actual process 11 can be used in the form of a database or a text file. If there is no actual data, the process model parameters can be automatically adjusted for use in the design.

  The process data interface 12 has a function of sequentially referring to the data of the distributed control system online. The process data interface 12 may be an OPC interface. The process data interface 12 extracts necessary data from the database in time series. In the offline state, the process data interface 12 serves as an interface for the data collection package, and allows data necessary for operation of the process model 10 to be referred to.

The data collection means 13 is a device that can reuse actual process operation data (operation amount, state amount, control amount) and simulator operation data immediately or in the future. The data collection means 13 is usually realized by a database.
The execution means 14 has an operation command function for operating the process model 10 in real time or in an accelerated state. The process model 10 performs the operation in response to the operation command.

  The steady operation state construction means 15 sets what is used as an input and a data amount for generating a steady state in the process model 10. Further, the steady operation state constructing means 15 generates the entire balance state from the structure of the process model 10. Normally, since data is insufficient, the steady operation state construction means 15 uses normal values for unset items or prompts for input. Furthermore, the evaluation function for calculation is given and the state where the value becomes small is automatically calculated including the static parameters of each model device. Also, when there is an imbalance in the amount given from the outside, the state and parameters are calculated in the same way.

  The adjusting means 16 calculates the dynamic / static parameters of each model device based on the actual process operation data and the simulation operation. In actual calculations, time-series error-related evaluation functions (such as weighted linearity, exponentiation, directionality evaluation, etc.) can be reduced, and parameters can be adjusted and searched using approximate evaluation of fluctuation shapes in time-series data. To do.

  To search for parameters, use a process model that is a strict model or a simplified model, and change the response to a measured value that is correlated with the manipulated variable that is correlated to the parameter to be searched and the response. Use as sensitivity (degree of influence). If there is a similar state in the past, give that state information for reference. In addition, a method of performing numerical analysis, a method of learning and estimating with a neural network according to an evaluation amount, a state, and the like are possible so as to reduce the evaluation.

The evaluation unit 17 calculates the evaluation used by the adjustment unit 16 or evaluates whether the balance of the data obtained by the data collection unit 13 is clearly incorrect or whether the fluctuation is too large. .
The database 18 stores the state used by the adjusting means 16 and the calculation result.
The adjustment of the first adjustment means in the claims is mainly performed by the steady operation state construction means 15, and the adjustment means 16 is also involved. Adjustment of the second adjustment means is performed by the adjustment means 16.

  FIG. 2 is a flowchart showing a procedure for adjusting parameters. The operation will be described according to the processing procedure of the flowchart.

(A1) Based on the information obtained via the process data interface 12 and the data collection unit 13, the adjustment unit 16 instructs the steady operation state construction unit 15 to generate the process state of the process model 10.

(A2) The steady operation state construction means 15 generates a plant state S1 in cooperation with the process model 10. If there is insufficient information (disturbance, prior error information, etc.), the information is set separately.

(A3) Based on the generated plant state S1 or the plant state S2 constructed in a later operation, the operation amount u to be added to the actual process is set to be added to the process model 10 as well. The operation amount added to the process model 10 is represented by us. The adjustment unit 16 instructs the execution unit 14 to perform a simulation, and the process model 10 and the actual process 11 are operated.

(A4) During operation, the process model 10, the actual process 11, the process data interface 12, the data collection unit 13, the execution unit 14, the adjustment unit 16, and the evaluation unit 17 are operating. Here, the evaluation means 17 monitors the process control amount y, the process model 10 control amount ys, and the process model 10 state amount xs, and moves to a parameter adjustment process A6 described later when the state greatly deviates. To do.

(A5) If the data of the process model 10 and the actual process 11 are within the reference values in the operation evaluation, the operation is continued, and if the parameter adjustment operation is not necessary, the operation is terminated.

(A6) In parameter adjustment, the adjustment means 16 cooperates with the database 18, and if reference data exists, the model parameters of the process model 10 are changed based on the parameters and adjustment methods from the past operation. If the adjustment has already been made in the plant state S2, the plant state S1, the past plant state S2, the adjustment parameter at that time, and the result are taken out from the data collecting means 13, and the parameter is changed again.

  In order to generate the state of the process model 10 based on the changed parameter, the process returns to the process A1 and the operation is continued. Assume that the estimated state is the plant state S2.

  The parameters changed and corrected in the processes A1 to A4 are stored in the database 18 together with the state at that time so that they can be reused. In addition, the time series information of the operation amount, the state amount, and the control amount are also saved.

FIG. 3 is a flowchart showing a specific processing procedure of the processing A2 of FIG. 2, and FIG. 4 is a processing procedure of the processing A6 of FIG.
FIG. 3 is a flowchart showing a procedure for creating a process model state. FIG. 4 is a flowchart showing the procedure for adjusting the parameters of the process model.

A processing procedure of the flowchart of FIG. 3 will be described.
(B1) The steady operation state construction means 15 gives the operation amount us, the control amount ys, and the disturbance d data to the process model 10 and instructs the calculation of the steady state.
(B2) The steady operation state construction means 15 determines the operation state of the process model 10 that minimizes the existing evaluation in the vicinity of the current parameter based on the given static parameter, the operation amount us, the control amount ys, and the disturbance d. Create

(B3) It is determined whether the parameter search is completed in the process model 10.
(B4) When the search ends, a steady state is created using the smallest evaluation amount so far.
(B5) If the search does not end, the process model 10 is searched so that the evaluation amount is reduced.

  In the adjustment of the static parameter, the static parameter is adjusted in consideration of the steady balance of the process model.

A processing procedure of the flowchart of FIG. 4 will be described.
(C1) The current operation amount us, control amount ys, disturbance d, and past data are searched for parameter calculation results in a state within a specified range.
(C2) The calculation error is compared in the process state obtained from the parameter obtained by the search, and the smaller one is stored.
(C3) It is determined whether there is parameter data within a specified range. If there is, return to process C2.

(C4) When there is no parameter data, it is determined whether there is any adjustable dynamic parameter data.
(C5) When there is no dynamic parameter data, the static parameter and the dynamic parameter having the smallest evaluation amount so far are stored in the database.
(C6) When there is dynamic parameter data, the dynamic parameter of the process model is changed so as to reduce the evaluation amount, and the process returns to the process C5.

  In the adjustment of the dynamic parameter, the dynamic parameter is adjusted in consideration of the dynamics of how the process moves. For example, when the dynamic parameter is changed and a signal is input to the process, the dynamic parameter is adjusted in consideration of response characteristics obtained by using the process model 10, the data collection unit 13, and the execution unit 14.

  As shown in the flowcharts of FIGS. 3 and 4, adjustment of the static parameters and adjustment of the dynamic parameters of the process model are performed separately.

  In addition, you may combine the function of a some component not only in the structure of an Example into one component. For example, the process model 10 and the steady operation state constructing means 15, the adjusting means 16 and the evaluating means 17, the data collecting means 13 and the database 18 may each be one component.

  In addition, the present invention can be used for correcting the output of a measuring instrument for a small object. By combining a plurality of sensors, information is accumulated, resulting in a highly accurate and reliable sensor.

  In the case of off-line, that is, when there is no actual process 11 and there is process operation data 19, the process operation can be accelerated and the parameter calculation can be performed at high speed. Quick adjustment is possible using actual operation data in the state.

  Even if some devices are replaced, much of the information accumulated so far may be left as it is, so that the model and state can be constructed efficiently.

  The process model itself may have a function of distinguishing dynamic parameters from static parameters.

It is a block diagram which shows one Example of this invention. It is the flowchart which showed the procedure which adjusts a parameter. It is the flowchart which showed the specific process sequence of process A2 of FIG. It is the flowchart which showed the specific process sequence of process A6 of FIG. It is the figure which showed the example of a structure of the model construction system used by the model prediction control in the past.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Process model 11 Actual process 13 Data collection means 14 Execution means 15 Steady operation state construction means 16 Adjustment means 17 Evaluation means

Claims (6)

In the process model adjustment method for adjusting a process model for simulating an actual process,
A method of adjusting a process model, characterized by separately adjusting static parameters and dynamic parameters of a process model based on data obtained from an actual process.   2. The process model adjustment method according to claim 1, wherein in the static parameter adjustment, the static parameter is adjusted in consideration of a steady balance of the process model.   2. The process model adjustment method according to claim 1, wherein in the adjustment of the dynamic parameter, the dynamic parameter is adjusted in consideration of dynamics of how the process moves.   2. The process model adjustment method according to claim 1, wherein in the dynamic parameter adjustment, the dynamic parameter is adjusted in consideration of a response characteristic when a signal is input to the process.   5. The process model adjustment method according to claim 1, wherein the process model has a function of distinguishing a dynamic parameter from a static parameter. In a process model adjustment device for adjusting a process model for simulating an actual process,
A first adjustment means for adjusting static parameters of the process model based on data obtained from an actual process;
A second adjusting means for adjusting a dynamic parameter of the process model based on data obtained from an actual process;
An apparatus for adjusting a process model, comprising:

Images