JP4307129B2 - Process state similar case search method and state prediction method, and computer-readable storage medium - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a state similar case search method and a state prediction method of a process for searching past cases similar to the operation state during the operation of a process such as a blast furnace, and a computer-readable storage medium.
[0002]
[Prior art]
Conventionally, when an abnormal state occurs in the operation of a process such as a blast furnace, an act of searching for a past case manually and responding to it has been performed. It is important to use past knowledge for future operational improvements, regardless of success or failure, but in the past there was no way to make full use of accumulated data, and it was common to rely on human memory. there were. For this reason, there is a problem that the action becomes a different opinion due to differences in personal experience.
[0003]
In contrast, case-based reasoning techniques are used in several fields. Patent Document 1 proposes a technique that applies case-based reasoning that solves a current problem based on past problem-solving cases. Patent Document 2 proposes a technique for supporting a systematization of specialized knowledge by proposing a tabular editor for case-based reasoning. In addition, Patent Documents 3 and 4 disclose a technique for predicting a process by creating a representative case by collecting a plurality of cases as a wastewater amount prediction system or a turbidity prediction system for a water purification plant process, creating a representative case. .
[0004]
[Patent Document 1]
JP-A-3-132826 [Patent Document 2]
JP-A-7-271588 [Patent Document 3]
JP 2001-288882 A [Patent Document 4]
[Patent Document 1] Japanese Patent Application Laid-Open No. 2002-119156
[Problems to be solved by the invention]
The process targeted by the present invention is a complex, non-linear and non-stationary process such as a blast furnace. Mathematical models that are symmetric with such processes have been developed but are not yet sufficient. For this reason, past operational knowledge and past cases are constructed and used as knowledge bases and case bases, but the actual situation is that efforts are being made to collect knowledge, construct case bases, and update them. In addition, if the update is not sufficient, there is a problem that the accuracy of search and prediction decreases.
[0006]
Further, in the methods disclosed in Patent Documents 3 and 4, since cases are collected, there is a problem that it is not possible to specify a time point similar to the current time point.
[0007]
Moreover, since it is necessary to create a case base for prediction from a plurality of cases in advance, the update is a problem. Case study departments have also been proposed, but problems remain in maintaining permanent performance.
[0008]
The present invention has been made in view of such circumstances, and can search past cases of complicated, nonlinear and non-stationary processes such as blast furnaces with high speed and high accuracy, can accurately carry out future prediction, and can be updated. The purpose is to provide an unnecessary online prediction method.
[0009]
[Means for Solving the Problems]
In order to achieve such an object, the state similar case search method according to the present invention is configured such that process data consisting of measured values of a plurality of process variables is stored at a constant cycle time together with a storage time and a storage number indicating the storage order. A method for retrieving past process state similar cases similar to the process state at a specified time from a time series database stored in
(B) for each of the plurality of process variables, a step of extracting a set of process data from the specified time Tk only to a past time point a predetermined number m, from the time series data base,
(B) Using the extracted series of process data as input data, a data value Xi that is a value of each of the input data is normalized using upper and lower limit values Ximax and Ximin given in advance, and then a predetermined quantization number and calculates the input data values Ixi quantized by integer after multiplying the steps of obtaining input data quantized,
(C) A series of steps including the step of storing the quantized input data and / or the specified time and / or the storage number of the time-series database together in the search table . Repetitively changing the time Tk to create a search table consisting of quantized past input data;
For each of the plurality of process variables, a step of extracting a set of process data from JP Teijikoku Ts which is separately specified to only past time predetermined number m from the time series data base,
The data value Xsi, which is the value of each of the extracted series of process data, is normalized using pre-supplied upper and lower limit values Xsimax and Xsimin, then multiplied by a predetermined quantization number, and then quantized to be quantized Calculating the obtained data value Ixsi to obtain quantized input data at the specific time Ts;
The quantized input time Ts is used as a key of the search table, and is compared with the past input data quantized in the search table for each corresponding data value to calculate the difference as a similarity. Searching for input data similar to the input data at the specific time Ts based on the similarity, and obtaining a storage number or storage time of a similar case from the similar input data;
Extracting similar process data from the time-series database based on the storage number or storage time of the similar case .
[0010]
State predicting method of the process according to the present invention, process data comprising a measure of the plurality of process variables, the sequence database when stored for each time constant period with storage number indicating the storage time and the storage order, designated A process similar to the process state at the specified time, and a past process state similar case search method,
(B) for each of the plurality of process variables, a step of extracting a set of process data from the specified time Tk only to a past time point a predetermined number m, from the time series data base,
(B) Using the extracted series of process data as input data, a data value Xi that is a value of each of the input data is normalized using upper and lower limit values Ximax and Ximin given in advance, and then a predetermined quantization number and calculates the input data values Ixi quantized by integer after multiplying the steps of obtaining input data quantized,
(C) A series of steps including the step of storing the quantized input data and / or the specified time and / or the storage number of the time-series database together in the search table . Repetitively changing the time Tk to create a search table consisting of quantized past input data;
For each of the plurality of process variables, a step of extracting a set of process data from JP Teijikoku Ts which is separately specified to only past time predetermined number m from the time series data base,
The data value Xsi, which is the value of each of the extracted series of process data, is normalized using pre-supplied upper and lower limit values Xsimax and Xsimin, then multiplied by a predetermined quantization number, and then quantized to be quantized Calculating the obtained data value Ixsi to obtain quantized input data at the specific time Ts;
The quantized input time Ts is used as a key of the search table, and is compared with the past input data quantized in the search table for each corresponding data value to calculate the difference as a similarity. Searching for input data similar to the input data at the specific time Ts based on the similarity, and obtaining a storage number or storage time of a similar case from the similar input data;
A similar time point is obtained based on the storage number or the storage time of the similar case, the process data of a predetermined time ΔT from the similar time point is obtained from the time series database, and the predetermined time from the specified specific time Ts. And obtaining a future state after ΔT .
[0011]
Another feature of the process state similar case search method and state prediction method according to the present invention is that the time series database targets a blast furnace process, and the process variable values are set as hot metal temperature, pulverized coal injection amount, solution loss. At least one or more are selected from carbon, heat flow ratio, charging pitch, Si amount, hot air temperature, furnace top temperature, and heat load.
[0012]
It is another feature of the state predicting how the process according to the present invention is to display the future state of the process, the step of displaying the time with a degree of reliability as likelihood corresponding to the similarity criterion prediction Is further provided .
[0013]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, preferred embodiments of the process state similar case search method and state prediction method and computer-readable storage medium of the present invention will be described.
[0014]
FIG. 1 is a block diagram showing a configuration of an apparatus used for carrying out a process similar case search method. Reference numeral 10 in FIG. 1 denotes a blast furnace process in which a plurality of various sensors for measuring temperature, pressure, component, position, and the like are installed.
[0015]
Reference numeral 20 in FIG. 1 is a blast furnace process measurement / control device, which presents time series data of various process information measured from the blast furnace process 10 to the process operator and, if necessary, the process process with the operator's intervention. Control operation is being performed.
[0016]
Reference numeral 30 in FIG. 1 denotes a process time series database, in which various process data measured by the measurement / control apparatus 20 are stored at regular time intervals. Actually, the storage cycle differs depending on the type of process data, but it is assumed here that the data is stored at a constant cycle. If not constant, long-cycle data can be converted to short-cycle data by holding a value in accordance with short-cycle data.
[0017]
FIG. 2 shows a data storage format of the process time series database 30. The stored process variables are hot metal temperature, pulverized coal injection (PCR), solution loss carbon (SLC), heat flow ratio, charging pitch, Si content, hot air temperature (Tf), furnace top gas temperature, thermal load ( Qloss). The storage number is a number indicating the order in which the time series data is stored. The storage DATE is the time when the time series data is stored, and is expressed in year / month / day / hour / minute / second.
[0018]
Reference numeral 40 in FIG. 1 denotes a search table creation unit, which creates a search table 50 used for similar case search from the process time series database 30. The search table 50 uses, as input variables, values obtained by quantizing the measured values of several process variables and the process state values calculated from the measured values and the values at the past several points. The quantization of the input variable is performed by converting the input variable value from 1 to N (N is a set value, for example, 25).
[0019]
In the case of the blast furnace process, each time point and the past two times for 9 variables of hot metal temperature, pulverized coal injection amount, solution loss carbon, heat flow ratio, charging pitch, Si amount, hot air temperature, furnace top gas temperature, and heat load. The index value of the search table 50 is created by quantizing each variable value into a value of 1 to 25, for example. As a variable to be searched in the search table 50, the date and time at that time or the storage number of the corresponding database is recorded.
[0020]
FIG. 3 shows an example of the search table 50. Here, the input variables are 27 variables (9 process variables, 3 times), and the number of quantization is 25. The search variable is a DB (database) storage number or a DB storage DATE. Input_1 to input_27 in FIG. 3 are input variables, each of which is related to nine variables such as hot metal temperature, pulverized coal injection amount, solution loss carbon, etc. (t), one hour before (t-1), two hours before (t-2) ) Is quantized.
[0021]
FIG. 4 shows a procedure for creating the search table 50. In step 41, the process time series database 30 is accessed in time series order, and data values, storage times, or storage numbers of predetermined input variables are collected from the process time series database 30.
[0022]
In step 42, the data value of the input variable is quantized. If the input data is Xi, the upper and lower limit values given in advance are Ximax and Ximin, and the quantization number is n, the quantized value Ixi of Xi is
Ixi = integerization (N * (Xi−Ximin) / (Ximax−Ximin) +1) (1)
(I = 1 ~ 27)
Calculated by Integerization means truncation after the decimal point. An example of the upper and lower limits of the input variable used for quantization is shown in FIG.
[0023]
In step 43, 27 quantized input data values and storage times or storage numbers are set in the search table 50.
[0024]
4 are repeated until a predetermined storage number or storage time in the process time series database 30.
[0025]
Returning to FIG. 1, reference numeral 60 in FIG. 1 denotes a similar case search unit, which searches for similar cases using the search table 50. In order to search the process state and the similar time point at the current time, the process data at the current time is fetched from the measurement / control apparatus 20, and similar cases are searched by the procedure shown in FIG. In step 61, process data at the current time point for which a similar case search is desired is obtained from the measurement / control apparatus 20.
[0026]
In step 62, the input data of the search table 50 is selected from the obtained process data, the quantization described above is performed, and 27 input values of the search table 50 are created.
[0027]
In step 63, the search table 50 is compared with the selected input data. Collation is performed by collating the search table 50 from top to bottom as a simple method. If there are items to be verified, record the number of stored DB storage numbers or DB storage DATEs as many as the verifications. In this case, the similarity of the case is set to 0. If there is no match in the search table 50 with the input data, the neighborhood 1 of the input data is searched again. The neighborhood data is first searched up to (+1, −1) before and after the quantized data value of each input data.
[0028]
For example, if the hot metal temperature quantized data is 13, the data in the range of 12 to 14 is the collation range. For each input variable, the range is expanded back and forth to search for one that satisfies all, and if there is, the DB storage number or DB storage DATE is recorded as a similar time point. In this case, the similarity is 1. Similarly, if there is nothing to search, the next neighborhood 2 is searched, and if there is something to be collated, the similar time point is recorded in the same manner, and the similarity is set to 2. A search is made up to the set neighborhood k, and if there is nothing to collate, there is no similar case.
[0029]
In step 64, similar process data is extracted from the time-series database 30 in order of a plurality of DB storage numbers or DB storage DATEs retrieved from the search table 50, and output to a screen or a file. Reference numeral 70 in FIG. 1 denotes a similar case display unit, which displays the search result.
[0030]
FIG. 7 shows a result example of a similar case search for a blast furnace. The correspondence of hot metal temperature, charging pitch, pulverized coal injection amount, solution loss carbon at the present time and similar time is shown. The upper row is the current case and the lower row is the similar case. From these results, the trend before and after the current case is similar to that before and after the similar case, and it is confirmed that the similar case search functions well.
[0031]
According to the method described above, it is very easy to update the search table 50. Although it is possible to execute it sequentially online, when a certain amount of update data is accumulated in the process time series database 30, a value is added to the search table 50 using the data in the procedure of FIG. Is possible. It is also easy to delete data before a certain point of time that cannot be used due to operational changes from the search table 50 by deleting the data of the storage date corresponding to the date and time, or the point in time. It can be carried out.
[0032]
In addition to searching for an event similar to the current state of the process, it is possible to search for an event similar to a specific point in time series database.
[0033]
In addition, a search corresponding to an input state that is not in the search table 50 can be performed by a neighborhood search of the search table 50, although the accuracy is somewhat reduced.
[0034]
FIG. 8 is a block diagram showing the configuration of an apparatus used for carrying out the process future state prediction method. In the following description, the same components as those in FIG. Reference numeral 65 denotes a state prediction unit, which searches the search table 50 for past cases similar to the time point at which future state prediction is desired by the same method as in FIG. 1, searches for similar times, and uses the future cases to Predict. The procedure will be described with reference to FIG. In step 61, process data at the current time point is obtained from the measurement / control apparatus 20.
[0035]
In step 62, input data of the search table 50 is selected from the obtained process data, quantization is performed, and input data of the search table 50 is created.
[0036]
In step 63, input data is set in the search table 50, and a plurality of stored similar time points are obtained. IR is used for the similarity at this time.
[0037]
In step 94, the process data is sequentially obtained from the time series database 30 using the future prediction time set from the similar time point, for example, the DB storage time up to 8 hours ahead as a key. Further, it is possible to obtain a plurality of sets of predicted future state values by repeating the same operation at a plurality of similar points. Individual future prediction graphs at a plurality of similar points, their averaged graphs, and fluctuation ranges are displayed on the screen. In addition, the reliability corresponding to the similarity of the case is displayed together with the display. For example, when the similarity is 0, the reliability is high (H), when the similarity is 1, the reliability is slightly high (SH), and when the similarity is 2, the reliability is normal (M). In the case of 3, the reliability is displayed as caution (L). Reference numeral 75 in FIG. 8 denotes a future state display section, which displays those future state predictions.
[0038]
Fig. 10 shows the future prediction results of the blast furnace. The actual value and predicted value of the hot metal temperature two hours ahead correspond well, indicating the effectiveness of this method.
[0039]
(Other embodiments)
The apparatus shown in FIGS. 1 and 8 can be realized by a computer, and software program codes for realizing the functions of the above-described embodiments are supplied to the computer, and the computer of the system or apparatus (CPU or MPU) is supplied. What was implemented by operating various devices according to the stored program is also included in the category of the present invention.
[0040]
In this case, the program code of the software itself realizes the functions of the above-described embodiments, and the program code itself constitutes the present invention. As a transmission medium for the program code, a communication medium (wired line or wireless line such as an optical fiber) in a computer network (LAN, WAN such as the Internet, wireless communication network, etc.) system for propagating and supplying program information as a carrier wave Etc.) can be used.
[0041]
Further, means for supplying the program code to the computer, for example, a recording medium storing the program code constitutes the present invention. As a recording medium for storing the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.
[0042]
【The invention's effect】
According to the present invention described in detail above, similar cases of complicated, non-linear and non-stationary processes can be searched at high speed, and if there is a time series database, similar search systematically by the method of the present invention. The device can be updated. As a result, the performance of the similarity search device can always be maintained with high accuracy. In addition, this makes it possible to always maintain the process state prediction performance with high accuracy. In addition, accurately predicting the future state of the process provides important guidance for determining future operational actions and greatly contributes to operational stability.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an apparatus used to implement a process similar case search method.
FIG. 2 is a diagram showing an example of a data storage format of a process time series database.
FIG. 3 is a diagram illustrating an example of a search table.
FIG. 4 is a flowchart illustrating a procedure for creating a search table.
FIG. 5 is a diagram illustrating upper and lower limit values used for quantization of an input variable.
FIG. 6 is a flowchart illustrating a procedure for searching for similar cases.
FIG. 7 is a diagram illustrating an example of a similar search result.
FIG. 8 is a block diagram showing the configuration of an apparatus used for carrying out a process future state prediction method.
FIG. 9 is a flowchart illustrating a state prediction procedure.
FIG. 10 is a diagram illustrating an example of a future state prediction result.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 10 Blast furnace process 20 Measurement / control apparatus 30 Process time-series database 40 Search multidimensional table creation part 50 Search multidimensional table 60 Similar case search part 65 State prediction part 70 Similar case display part 75 Future state display part

Claims (6)

Process data consisting of the measured values of each of the plurality of process variables, the sequence database when stored for each time constant period with storage number indicating the storage time and the storage order, past similar to the process state of the specified time A method for searching for process state-like cases,
(B) for each of the plurality of process variables, a step of extracting a set of process data from the specified time Tk only to a past time point a predetermined number m, from the time series data base,
(B) Using the extracted series of process data as input data, a data value Xi that is a value of each of the input data is normalized using upper and lower limit values Ximax and Ximin given in advance, and then a predetermined quantization number and calculates the input data values Ixi quantized by integer after multiplying the steps of obtaining input data quantized,
(C) A series of steps including the step of storing the quantized input data and / or the specified time and / or the storage number of the time-series database together in the search table . Repetitively changing the time Tk to create a search table consisting of quantized past input data;
For each of the plurality of process variables, a step of extracting a set of process data from JP Teijikoku Ts which is separately specified to only past time predetermined number m from the time series data base,
The data value Xsi, which is the value of each of the extracted series of process data, is normalized using pre-supplied upper and lower limit values Xsimax and Xsimin, then multiplied by a predetermined quantization number, and then quantized to be quantized Calculating the obtained data value Ixsi to obtain quantized input data at the specific time Ts;
The quantized input time Ts is used as a key of the search table, and is compared with the past input data quantized in the search table for each corresponding data value to calculate the difference as a similarity. Searching for input data similar to the input data at the specific time Ts based on the similarity, and obtaining a storage number or storage time of a similar case from the similar input data;
A process state similar case retrieval method comprising: extracting similar process data from the time series database based on a storage number or a storage time of the similar case.   The time series database targets a blast furnace process, and the process variable value is at least 1 from the hot metal temperature, pulverized coal injection amount, solution loss carbon, heat flow ratio, charging pitch, Si amount, hot air temperature, furnace top temperature, and thermal load. The process state similarity case search method according to claim 1, wherein one or more are selected. Process data consisting of the measured values of each of the plurality of process variables, the sequence database when stored for each time constant period with storage number indicating the storage time and the storage order, past similar to the process state of the specified time A method for searching for process state-like cases,
(B) for each of the plurality of process variables, a step of extracting a set of process data from the specified time Tk only to a past time point a predetermined number m, from the time series data base,
(B) Using the extracted series of process data as input data, a data value Xi that is a value of each of the input data is normalized using upper and lower limit values Ximax and Ximin given in advance, and then a predetermined quantization number and calculates the input data values Ixi quantized by integer after multiplying the steps of obtaining input data quantized,
(C) A series of steps including the step of storing the quantized input data and / or the specified time and / or the storage number of the time-series database together in the search table . Repetitively changing the time Tk to create a search table consisting of quantized past input data;
For each of the plurality of process variables, a step of extracting a set of process data from JP Teijikoku Ts which is separately specified to only past time predetermined number m from the time series data base,
The data value Xsi, which is the value of each of the extracted series of process data, is normalized using pre-supplied upper and lower limit values Xsimax and Xsimin, then multiplied by a predetermined quantization number, and then quantized to be quantized Calculating the obtained data value Ixsi to obtain quantized input data at the specific time Ts;
The quantized input time Ts is used as a key of the search table, and is compared with the past input data quantized in the search table for each corresponding data value to calculate the difference as a similarity. Searching for input data similar to the input data at the specific time Ts based on the similarity, and obtaining a storage number or storage time of a similar case from the similar input data;
A similar time point is obtained based on the storage number or the storage time of the similar case, the process data of a predetermined time ΔT from the similar time point is obtained from the time series database, and the predetermined time from the specified specific time Ts. And a step of obtaining a future state after ΔT .   The time series database targets a blast furnace process, and the process variable value is at least 1 from the hot metal temperature, pulverized coal injection amount, solution loss carbon, heat flow ratio, charging pitch, Si amount, hot air temperature, furnace top temperature, and thermal load. The process state prediction method according to claim 3, wherein one or more are selected. 5. The process according to claim 3 , further comprising a step of displaying a future state of the process, and displaying a reliability corresponding to the similarity criterion as a probability of prediction at that time. State prediction method. A process for causing a computer to execute each process in the step of the process state similarity case search method according to claim 1 or 2 , or the step of the process state prediction method according to any one of claims 3 to 5. A computer-readable storage medium characterized by recording a program.

Images