CN112859793B - Industrial production process dynamic time delay identification method based on improved sliding time window - Google Patents
Industrial production process dynamic time delay identification method based on improved sliding time window Download PDFInfo
- Publication number
- CN112859793B CN112859793B CN202110171300.6A CN202110171300A CN112859793B CN 112859793 B CN112859793 B CN 112859793B CN 202110171300 A CN202110171300 A CN 202110171300A CN 112859793 B CN112859793 B CN 112859793B
- Authority
- CN
- China
- Prior art keywords
- time delay
- input variable
- dynamic
- time window
- variable
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 85
- 238000009776 industrial production Methods 0.000 title description 5
- 238000004519 manufacturing process Methods 0.000 claims abstract description 34
- 230000008859 change Effects 0.000 claims abstract description 23
- 230000003068 static effect Effects 0.000 claims abstract description 17
- 230000008569 process Effects 0.000 claims description 26
- 238000005259 measurement Methods 0.000 claims description 14
- 239000001273 butane Substances 0.000 claims description 8
- IJDNQMDRQITEOD-UHFFFAOYSA-N n-butane Chemical compound CCCC IJDNQMDRQITEOD-UHFFFAOYSA-N 0.000 claims description 8
- OFBQJSOFQDEBGM-UHFFFAOYSA-N n-pentane Natural products CCCCC OFBQJSOFQDEBGM-UHFFFAOYSA-N 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 6
- 230000007246 mechanism Effects 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 4
- 238000010992 reflux Methods 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract 1
- 230000008901 benefit Effects 0.000 description 5
- 230000001934 delay Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000000691 measurement method Methods 0.000 description 3
- 230000035772 mutation Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000006477 desulfuration reaction Methods 0.000 description 1
- 230000023556 desulfurization Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
- G05B19/41875—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by quality surveillance of production
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32368—Quality control
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Manufacturing & Machinery (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses an industrial process dynamic time delay identification method based on an improved sliding time window, which adopts a static delimitation-dynamic updating strategy to estimate dynamic time delay, firstly adopts a static FCA method to estimate total time delay so as to determine a dynamic time delay change range, and then carries out time delay estimation on each time window based on the improved sliding time window to mine the dynamic time delay characteristic among variables. The method and the device more deeply excavate the time delay characteristics among the variables, realize more accurate time registration among the variables and are beneficial to improving the prediction precision of the output variables.
Description
Technical Field
The invention relates to the technical field of complex industrial process modeling and control, in particular to an industrial production process dynamic time delay identification method based on an improved sliding time window.
Background
The key quality variable in the complex industrial process is an important variable for measuring the production effect of the product, and the accurate prediction and control of the key quality variable are the basis of the optimized operation of the industrial process. However, these variables often have the characteristics of high measurement difficulty and long measurement period, and therefore, a soft measurement method based on data driving is often adopted to realize prediction and control of the variables. However, in an actual industrial process, input and output variable data of the production process are not matched in time due to the time required for transmission of signals and raw materials, and the output often lags behind the input. At the moment, the prediction model established by the current input and output variable data cannot better explain the process characteristics, and the prediction and control of key quality variables (output variables) in the production process of the product are greatly limited. Therefore, in order to better explain the industrial process, it is important to consider the time lag of the variables in the process of realizing the prediction and control of the output variables.
There are many researchers working on the problem of estimating the time delay between industrial process variables, and these methods mainly include two types: firstly, the time delay among variables is determined according to an industrial process mechanism and expert experience, but the time delay estimated by the method is low in accuracy, secondly, the time delay estimation is carried out by adopting a data-driven method based on process data, the main idea is to search the time delay of the variable with the maximum correlation with the output variable, and the method is high in precision and persuasive, so that the method is widely applied. Based on the second method, a scholars determines the time delay by calculating the pearson correlation coefficient of different time delay input variables and output variables, however, the pearson correlation coefficient measures the linear correlation between the variables, and is not suitable for the nonlinear industrial process. Therefore, some inter-variable nonlinear correlation measurement methods are also used for delay estimation, such as a mutual information method, but due to the high computational complexity of the mutual information method, after a tradeoff between the computational complexity and the accuracy of delay estimation, an FCA (Fuzzy Analysis, FCA) Fuzzy Curve method is selected for time delay estimation between variables. However, these methods are all to estimate the total delay of the variables, neglecting the dynamic delay characteristics of the variables, but in the actual industrial process, the transmission of the material and the transmission of the signal are fluctuated, so the time delay between the variables is also dynamically changed, and the estimation of the dynamic delay between the variables can reflect the actual production process better.
Therefore, aiming at the dynamic time delay characteristics among the actual industrial process variables, the invention provides a strategy based on static delimitation-dynamic update estimation to mine the dynamic time delay characteristics among the industrial process variables, realize accurate time registration among the input and output variables and lay a foundation for real-time prediction and control of later-stage key quality variables.
Disclosure of Invention
Technical problem to be solved
Based on the problems, the invention provides an industrial production process dynamic time delay identification method based on an improved sliding time window, which is used for mining the dynamic time delay characteristics among industrial process variables based on a static delimitation-dynamic updating estimation strategy, realizing accurate time registration among input and output variables and solving the problem that the dynamic time delay of the variables is difficult to determine and further influences the prediction precision.
(II) technical scheme
Based on the technical problem, the invention provides an industrial production process dynamic time delay identification method based on an improved sliding time window, which comprises the following steps:
s1, acquiring input variable and output variable data of the industrial process of the debutanizer under the condition of a fixed sampling period, and determining the maximum time delay T of each input variable relative to the output variable according to the internal mechanism and expert experience of the industrial processmax(ii) a The input variables comprise the tower top temperature, the tower top pressure, the tower top reflux quantity, the tower top product outflow quantity, the tower plate temperature of the sixth layer and the tower bottom temperature, and the output variables comprise the tower bottom butane concentration;
s2, for the input variable xiAnd performing variable time reconstruction on the output variable y to obtain an input variable set X (t) ([ X ]) with time delay1(t),X2(t),…,Xm(t)]The reconstruction of each input variable is as follows:
Xi(t)＝[xi(t),xi(t-1),…,xi(t-λ),…,xi(t-Tmax)],
wherein x isi(t) is the original input variable at time t, xi(T- λ) is an input variable whose time delay at time T is λ, i is 1,2, …, m is the number of input variables, λ is 0,1, …, T is 0max；
S3, static delimitation: the FCA fuzzy curve method is used for carrying out static time delay estimation on each input variable, the obtained time delay is the total time delay d of each input variable, and therefore the dynamic time delay change range [ d ] is obtainedlow,dup]；
S4, dynamic estimation: observing the change condition of the dynamic time delay under different time window sizes, and further determining the initial value L of the sliding time windowiniAnd maximum value Lmax: when the time delay shows obvious dynamic change characteristics and the time delay exceeding 80 percent does not exceed the change range of the dynamic time delay, taking the window size at the moment as the initial value of a sliding time window; similarly, when the change of the time delay tends to be flat, the window size at the moment is taken as the maximum value of the sliding time window;
s5, estimating the time delay of each input variable of the current sliding time window by adopting the FCA fuzzy curve method to obtain each inputTime delay d of current sliding time window of variablecur；
S6, respectively judging whether the time delay of the current sliding time window of each input variable exceeds the dynamic time delay variation range corresponding to each input variable, if so, determining whether the current sliding time window of each input variable exceeds the dynamic time delay variation range corresponding to each input variable
Increasing the size of the sliding time window and returning to the step S5, otherwise, performing the step S7;
s7, respectively judging the time delay d of the current sliding time window of each input variablecurTime delay d corresponding to the previous sliding time windowpreWhether the absolute value of the difference of (d) exceeds a sudden change threshold dmIf | dcur-dpre|＞dmThen the size of the sliding time window continues to be increased, and the process proceeds to step S8; otherwise, go to step S9;
s8, respectively judging whether the current sliding time window size of each input variable exceeds the maximum value of the sliding time window, if not, returning to the step S5; if yes, replacing the time delay of the current sliding time window of the input variable with the total time delay corresponding to the input variable, and entering step S9;
s9, storing the time delay of the current sliding time window of each input variable, judging whether the dynamic time delay estimation of all the sliding time windows is finished, if not, moving the sliding time windows one step to the next moment, returning to the step S5, otherwise, entering the step S10;
and S10, performing soft measurement modeling on the output variable by using each input variable containing dynamic time delay, and realizing the prediction of the output variable.
Further, the specific process of performing the delay estimation on each input variable by the FCA fuzzy curve method in steps S3 and S5 includes the following steps:
s3.1, input and output variable set { xiY, i ═ 1, … m }, and at time t the variable xiThe gaussian fuzzy membership function of (a) is:
wherein k is 0.2 times of the value range of the variable xi and is in the range of { xi(t), y (t), where i is 1, … m }, phiit(xi)＝1；
S3.2, carrying out centroid defuzzification on the input variables with different time delays to obtain a fuzzy curve value of the variable xi when the time delay is lambda, namely
S3.3, finding out fuzzy curve value
Taking the time delay corresponding to the maximum value and the minimum value, and respectively recording the time delay as lambda1,iAnd λ2,iFinally, the variable x is obtainediOf optimal delay size di：
di＝λ1,i-λ2,i。
Further, the soft measurement modeling model described in step S10 is a correlation vector machine model.
Further, before performing soft measurement modeling in step S10, the method further includes: counting the number of windows occupied by each time delay of each input variable, acquiring a dynamic time delay identification result of the input variable for verification, and calculating an asymmetric dependence coefficient of each input variable including the dynamic time delay on an output variable for further verification.
Further, the calculation method of the asymmetric dependency coefficient of each input variable including the dynamic time delay on the output variable is as follows:
wherein, H (x)1,…,xn) And H (y) represents the magnitude of the information entropy of variables x and y, respectively, H (x)1,…,xnAnd y) represents the joint information entropy of the variables x and y.
The invention also discloses an industrial process dynamic time delay identification system based on the improved sliding time window, which comprises the following steps:
at least one processor; and at least one memory communicatively coupled to the processor, wherein:
the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method.
A non-transitory computer readable storage medium storing computer instructions that cause the computer to perform the method is also disclosed.
(III) advantageous effects
The technical scheme of the invention has the following advantages:
(1) the method changes the fixed window size in the traditional method into the variable window size, adopts a static delimitation-dynamic updating strategy to estimate the dynamic time delay, can better mine the dynamic time delay characteristic of the variable on the premise of ensuring the effectiveness, realizes more accurate time registration among the variables, and has important significance for improving the prediction precision and the monitoring level of the key quality variable;
(2) the invention applies the FCA-based time delay estimation method to the overall time delay estimation of the whole process and the dynamic time delay estimation of each time window, and fully utilizes the advantage that the FCA method has lower algorithm complexity on the premise of ensuring the estimation precision.
Drawings
The features and advantages of the present invention will be more clearly understood by reference to the accompanying drawings, which are illustrative and not to be construed as limiting the invention in any way, and in which:
FIG. 1 is a flow chart of a dynamic delay identification method based on an improved sliding time window according to the present invention;
FIG. 2 is a simplified schematic diagram of a debutanizer process according to an embodiment of the present invention;
FIG. 3 is a debutanizer process input variable x for an embodiment of the present invention1And dynamic time delay variation between output variablesCurve transformation;
FIG. 4 is a debutanizer process input variable x for an embodiment of the present invention1The number of windows contained in the time delay different from that of the output variable;
FIG. 5 is a process variable x for a debutanizer column according to an embodiment of the present invention1A time window size change curve identified by the dynamic time delay of the output variable;
FIG. 6 is a graph of the predicted effect of butane concentration on the test set based on the RVM-DTDE method of the present invention.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
An industrial process dynamic time delay identification method based on an improved sliding time window is shown in fig. 1, and in the embodiment, an actual industrial process, namely a debutanizer process is taken as an example, and the debutanizer process is an important component of a desulfurization and naphtha decomposition device. In this process, one of the main tasks is to reduce the butane concentration at the bottom of the debutanizer column. The simplified process flow diagram of the whole process is shown in fig. 2. The method comprises the following steps:
s1, acquiring input variable and output variable data of the industrial process of the debutanizer under the condition of a fixed sampling period, and determining the maximum time delay T of each input variable relative to the output variable according to the internal mechanism and expert experience of the industrial processmax；
The debutanizer process comprises 7 input variables which can be measured in real time and an difficultly-measured mass variable, namely the output variable of the concentration of butane at the bottom of the tower, and the whole process is sampled once every 6 minutes to obtain 2394 continuous samples. Table 1 gives a detailed description of these variables, and for the sake of simplicity, two bottoms temperature variables x6，x7Averaged and combined into one variable.
TABLE 1
Setting the maximum time delay T between input and output variables according to the process mechanism and the related literature introductionmax19 sampling periods.
S2, for the input variable xiAnd performing variable time reconstruction on the output variable y to obtain an input variable set X (t) ([ X ]) with time delay1(t),X2(t),…,Xm(t)]The reconstruction method of each input variable is shown in formula (1):
Xi(t)＝[xi(t),xi(t-1),…,xi(t-λ),…,xi(t-Tmax)] (1)
wherein x isi(t) is the original input variable at time t, xi(T- λ) is an input variable whose time delay at time T is λ, i is 1,2, …, m is the number of input variables, λ is 0,1, …, T is 0max，TmaxFor small to large delays, TmaxFor maximum delay, in this embodiment, according to the set maximum delay, the original input variable is then expanded to the input variable { X with different delaysi(t-λ),λ＝0,1,…,19,i＝1，…，6}
S3, static delimitation: the FCA method is used for carrying out static time delay estimation on each input variable, the obtained time delay is the total time delay d of each input variable, and therefore the dynamic time delay change range [ d ] is obtainedlow,dup]；
The specific process of performing static delay estimation on each input variable by using the FCA method in step S3 is as follows:
s3.1, input and output variable set { x is assumediY, i ═ 1, … m }, and at time t the variable xiThe definition of the Gaussian fuzzy membership function is shown as a formula (2), wherein k is a variable xi0.2 times the range of values, and is in { xi(t), y (t), where i is 1, … m }, phiit(xi)＝1。
S3.2 for formula (1)Inputting variables with different time delays, and performing centroid defuzzification on the variables by using a formula (3) to obtain a variable xiFuzzy curve value at time delay λ:
s3.3, finding out fuzzy curve value
Taking the time delay corresponding to the maximum value and the minimum value, and respectively recording the time delay as lambda1,iAnd λ2,iFinally, the variable x is obtainediOf optimal delay size diAs shown in equation (4):
di＝λ1,i-λ2,i (4)
s4, dynamic estimation: observing the change condition of the dynamic time delay under different time window sizes, and further determining the initial value L of the sliding time windowiniAnd maximum value Lmax: when the time delay shows obvious dynamic change characteristics and the time delay exceeding 80 percent does not exceed the change range of the dynamic time delay, taking the window size at the moment as the initial value of a sliding time window; similarly, when the change of the time delay tends to be flat, the window size at the moment is taken as the maximum value of the sliding time window; the obvious dynamic change characteristic indicates that the time delay is dynamically changed; the variation range of the time delay exceeding 80% does not exceed the dynamic time delay is set to 80% according to experience;
in this embodiment, the total delay and the dynamic delay variation range of each input variable obtained in steps S3 and S4, and the initial value and the maximum value of the sliding time window size are as shown in table 2 below, where the delay sizes of different windows may vary around the total delay value, but the variation is not severe, so the variation range in this embodiment is determined to be 10;
TABLE 2
S5, dynamic estimation: estimating the time delay of each input variable of the current time window by adopting the FCA method to obtain the time delay d of the current time window of each input variablecur；
S6, respectively judging whether the time delay of the current time window of each input variable exceeds the dynamic time delay variation range corresponding to each input variable, if so, determining whether the current time window of each input variable exceeds the dynamic time delay variation range corresponding to each input variable
Increasing the size of the sliding time window and returning to step S5, otherwise proceeding to step S7;
in this embodiment, if yes, the current window size is increased by 50;
s7, respectively judging the time delay d of the current time window of each input variablecurTime delay d corresponding to the previous time windowpreWhether the absolute value of the difference of (d) exceeds a sudden change threshold dmIf | dcur-dpre|＞dmIf yes, continue to increase the size of the sliding time window, and go to step S8; otherwise, go to step S9;
in this embodiment, the time delay mutation threshold of each variable is set to 5, and if the time delay mutation threshold exceeds the time delay mutation threshold, the size of the current window is increased by 50;
s8, respectively judging whether the current time window size of each input variable exceeds the maximum value of the sliding time window size, if not, returning to the step S5; if yes, replacing the time delay of the current time window of the input variable with the total time delay corresponding to the input variable, and entering step S9;
s9, storing the time delay of the current time window of each input variable, judging whether the dynamic time delay estimation of all the time windows is finished, if not, sliding the time window one step to the next moment, returning to the step S5, otherwise, entering the step S10;
and S10, performing soft measurement modeling on the output variable by using each input variable containing dynamic time delay, and realizing the prediction of the output variable.
Before soft measurement modeling is carried out, the invention also counts the number of windows occupied by each time delay of the current variable, obtains the dynamic time delay identification result of the input variable for verification, and calculates the Asymmetric Dependence Coefficient (ADC) of each input variable containing the dynamic time delay on the output variable for further verification;
the calculation mode of the asymmetric dependence coefficient of each input variable including the dynamic time delay on the output variable is as follows:
assuming that there are two variables x, y, the formula for calculating the asymmetric dependency coefficients of the two variables is shown in formula (5), where H (x) and H (y) represent the magnitude of the information entropy of the variables x and y, respectively, and H (x, y) represents the joint information entropy of the variables x and y.
If the condition of a single input variable is expanded to the condition of a high-dimensional input variable, the asymmetric dependency coefficient calculation formula of the multi-dimensional input variable and the output variable becomes as shown in formula (6):
it can be seen from the formula for defining the asymmetric dependency coefficient, which measures the proportion of the information amount of the output variable contained in the input variable, and the value of the input variable is closer to 1 as the input variable contains more information of the output variable. That is, when the identified dynamic delay between the input variable and the output variable is more accurate, the input variable can describe the output variable more, and the value of the asymmetric dependency coefficient is closer to 1, so that the identified dynamic delay result can be further verified through the magnitude of the asymmetric dependency coefficient value.
In this embodiment, the result of identifying the dynamic delay of each input variable is shown in table 3, where one input variable x is used1For illustration, the method of the present invention is used to input variable x1And an output variableTo obtain a variable x1The dynamic delay variation of (2) is shown in FIG. 3, and the input variable x is counted1The number of windows included in each time delay is shown in fig. 4, and the number of windows included in the time delays 13,14 and 15 is obviously greater than that of other time delays, so that the input variable x is determined1Including a dynamic delay of x1(t-13)、x1(t-14)、x1(t-15); dynamic time delay determination mode of other input variables is the same as that of input variable x1(ii) a Fig. 5 shows the size change of the sliding time window in the dynamic delay estimation process, which is mainly used to show how the time window changes when a sudden delay change or a deviation from the overall delay is encountered in the dynamic delay estimation process.
TABLE 3
In this embodiment, a Relevance Vector Machine (RVM) model is selected as the soft measurement model, and the prediction accuracy of the model adopts an Asymmetric Dependent Coefficient (ADC), a Root Mean Square Error (RMSE), and a decision coefficient (R)2) And (5) carrying out measurement.
Meanwhile, in order to compare the effectiveness of the present invention, the prediction accuracy of the butane concentration was compared between the correlation vector machine model (RVM-DTDE) based on the dynamic delay estimation, the correlation vector machine model (RVM-NTDE) based on the non-delay estimation, and the correlation vector machine model (RVM-STDE) based on the static delay estimation, and the results are shown in table 4.
TABLE 4
The simulation results show that the time delay estimation is carried out on the process variables of the debutanizer, the information that the input variables comprise the output variables can be more fully mined, and the prediction precision of the soft measurement model on the concentration of the debutanizer is effectively improved. Compared with a soft measurement model RVM-STDE based on static time delay estimation, the dynamic time delay characteristic of the process can be more fully excavated through dynamic time delay estimation compared with static time delay estimation, and higher prediction accuracy is obtained. Fig. 6 is a prediction curve diagram of the soft measurement method based on RVM-DTDE for butane concentration, and it can be seen that the method can better fit the actual output value to the prediction value of butane concentration overall, and the validity of the dynamic delay characteristic mining between actual process variables based on the dynamic delay identification method provided by the invention is verified.
Compared with the traditional time delay identification method based on the static sliding time window, the method is based on two assumption conditions of 1) and 2):
1) the dynamic time delay of the variables should vary around the overall time delay;
2) the time delay from the previous window of the variable to the current window does not jump.
Finally, it should be noted that the above-described method can be converted into software program instructions, and can be implemented by using a control system including a processor and a memory, or by using computer instructions stored in a non-transitory computer-readable storage medium. The integrated unit implemented in the form of a software functional unit may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute some steps of the methods according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.
In summary, the method for identifying the dynamic time delay of the industrial process based on the improved sliding time window has the following advantages:
(1) the method changes the fixed window size in the traditional method into the variable window size, adopts a static delimitation-dynamic updating strategy to estimate the dynamic time delay, can better mine the dynamic time delay characteristic of the variable on the premise of ensuring the effectiveness, realizes more accurate time registration among the variables, and has important significance for improving the prediction precision and the monitoring level of the key quality variable;
(2) the invention applies the FCA-based time delay estimation method to the overall time delay estimation of the whole process and the dynamic time delay estimation of each time window, and fully utilizes the advantage that the FCA method has lower algorithm complexity on the premise of ensuring the estimation precision.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.
Claims (8)
1. An industrial process dynamic time delay identification method based on an improved sliding time window is characterized by comprising the following steps:
s1, acquiring input variable and output variable data of the industrial process of the debutanizer under the condition of a fixed sampling period, and determining each input variable x according to the internal mechanism and expert experience of the industrial processiMaximum time delay T relative to output variable ymax；
S2, for the input variable xiAnd performing variable time reconstruction on the output variable y to obtain an input variable set X (t) ([ X ]) with time delay1(t),X2(t),…,Xm(t)]The reconstruction of each input variable is as follows:
Xi(t)＝[xi(t),xi(t-1),…,xi(t-λ),…,xi(t-Tmax)],
wherein x isi(t) is the original input variable at time t, xi(T- λ) is an input variable whose time delay at time T is λ, i is 1,2, …, m is the number of input variables, λ is 0,1, …, T is 0max；
S3, static delimitation: static time delay estimation is carried out on each input variable by using an FCA fuzzy curve method to obtain timeDelaying to the total delay d of each input variable to obtain the dynamic delay variation range dlow,dup]；
S4, dynamic estimation: observing the change condition of the dynamic time delay under different time window sizes, and further determining the initial value L of the sliding time windowiniAnd maximum value Lmax: when the time delay shows obvious dynamic change characteristics and the time delay exceeding 80 percent does not exceed the change range of the dynamic time delay, taking the window size at the moment as the initial value of a sliding time window; similarly, when the change of the time delay tends to be flat, the window size at the moment is taken as the maximum value of the sliding time window;
s5, estimating the time delay of each input variable of the current sliding time window by adopting the FCA fuzzy curve method to obtain the time delay d of the current sliding time window of each input variablecur；
S6, respectively judging whether the time delay of the current sliding time window of each input variable exceeds the dynamic time delay variation range corresponding to each input variable, if so, determining whether the current sliding time window of each input variable exceeds the dynamic time delay variation range corresponding to each input variable
Increasing the size of the sliding time window and returning to step S5, otherwise proceeding to step S7;
s7, respectively judging the time delay d of the current sliding time window of each input variablecurTime delay d corresponding to the previous sliding time windowpreWhether the absolute value of the difference of (d) exceeds a sudden change threshold dmIf | dcur-dpre|＞dmThen the size of the sliding time window continues to be increased, and the process proceeds to step S8; otherwise, go to step S9;
s8, respectively judging whether the current sliding time window size of each input variable exceeds the maximum value of the sliding time window, if not, returning to the step S5; if yes, replacing the time delay of the current sliding time window of the input variable with the total time delay corresponding to the input variable, and entering step S9;
s9, storing the time delay of the current sliding time window of each input variable, judging whether the dynamic time delay estimation of all the sliding time windows is finished, if not, moving the sliding time windows one step to the next moment, returning to the step S5, otherwise, entering the step S10;
and S10, performing soft measurement modeling on the output variable by using each input variable containing dynamic time delay, and realizing the prediction of the output variable.
2. The improved sliding time window based industrial process dynamic time delay identification method according to claim 1, wherein the soft measurement modeling model in step S10 is a correlation vector machine model.
3. The method for identifying the dynamic time delay of the industrial process based on the improved sliding time window of claim 1, wherein before the step S10 of performing the soft measurement modeling, the method further comprises: counting the number of windows occupied by each time delay of each input variable, acquiring a dynamic time delay identification result of the input variable for verification, and calculating an asymmetric dependence coefficient of each input variable including the dynamic time delay on an output variable for further verification.
4. The improved sliding time window-based industrial process dynamic time delay identification method according to claim 3, wherein the asymmetric dependency coefficients of the input variables including the dynamic time delay on the output variables are calculated in a manner that:
wherein, H (x)1,...,xn) And H (y) represents the magnitude of the information entropy of variables x and y, respectively, H (x)1,…,xnAnd y) represents the joint information entropy of the variables x and y.
5. The improved sliding time window based industrial process dynamic time delay identification method of claim 1, wherein the input variables comprise tower top temperature, tower top pressure, tower top reflux, tower top product outflow, sixth layer tower plate temperature, and tower bottom temperature, and the output variables comprise tower bottom butane concentration.
6. The method for identifying the dynamic time delay of the industrial process based on the improved sliding time window as claimed in claim 1, wherein the dynamic time delay of each input variable is changed around the total time delay, and the time delay from the previous sliding time window to the current sliding time window of each input variable does not change suddenly.
7. An industrial process dynamic time delay identification system based on an improved sliding time window, comprising:
at least one processor; and at least one memory communicatively coupled to the processor, wherein:
the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of claims 1 to 6.
8. A non-transitory computer-readable storage medium storing computer instructions that cause a computer to perform the method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110171300.6A CN112859793B (en) | 2021-02-04 | 2021-02-04 | Industrial production process dynamic time delay identification method based on improved sliding time window |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110171300.6A CN112859793B (en) | 2021-02-04 | 2021-02-04 | Industrial production process dynamic time delay identification method based on improved sliding time window |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112859793A CN112859793A (en) | 2021-05-28 |
| CN112859793B true CN112859793B (en) | 2022-05-10 |
Family
ID=75989053
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110171300.6A Active CN112859793B (en) | 2021-02-04 | 2021-02-04 | Industrial production process dynamic time delay identification method based on improved sliding time window |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112859793B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112270449A (en) * | 2020-11-04 | 2021-01-26 | 国网(苏州)城市能源研究院有限责任公司 | Industrial system time delay determination and controlled quantity prediction method based on time correlation |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6577960B1 (en) * | 2000-07-13 | 2003-06-10 | Simmonds Precision Products, Inc. | Liquid gauging apparatus using a time delay neural network |
| US7610397B2 (en) * | 2005-02-28 | 2009-10-27 | International Business Machines Corporation | Method and apparatus for adaptive load shedding |
| RU2541848C1 (en) * | 2013-12-03 | 2015-02-20 | Евгений Александрович Сухарев | Adaptive control system |
| CN105205224B (en) * | 2015-08-28 | 2018-10-30 | 江南大学 | Time difference Gaussian process based on fuzzy curve analysis returns soft-measuring modeling method |
| CN106156434B (en) * | 2016-07-11 | 2020-03-17 | 江南大学 | Sliding window time difference-Gaussian process regression modeling method based on local time lag reconstruction |
| CN106845095A (en) * | 2017-01-10 | 2017-06-13 | 上海第二工业大学 | A kind of recognition methods of 2-KLG industrial fermentation processes metabolic activity critical stage |
| CN107273633B (en) * | 2017-06-29 | 2018-03-27 | 中南大学 | Varying delay method of estimation and flow time lag method of estimation is hydrocracked between multiple operation |
| CN108650065B (en) * | 2018-03-15 | 2021-09-10 | 西安电子科技大学 | Window-based streaming data missing processing method |
-
2021
- 2021-02-04 CN CN202110171300.6A patent/CN112859793B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112270449A (en) * | 2020-11-04 | 2021-01-26 | 国网(苏州)城市能源研究院有限责任公司 | Industrial system time delay determination and controlled quantity prediction method based on time correlation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112859793A (en) | 2021-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Johnson et al. | A predictive view of the detection and characterization of influential observations in regression analysis | |
| CN111563706A (en) | Multivariable logistics freight volume prediction method based on LSTM network | |
| CN108241901B (en) | Transformer early warning evaluation method and device based on prediction data | |
| CN109800454B (en) | Coalbed methane collection method and system based on Kalman filtering and storage medium | |
| CN112116138A (en) | Power system prediction state estimation method and system based on data driving | |
| CN107423503A (en) | The GPR modelings estimated based on the slow signature analysis of core and time lag | |
| CN111638707A (en) | Intermittent process fault monitoring method based on SOM clustering and MPCA | |
| CN104766137A (en) | Network security posture prediction method based on evidence theory | |
| CN108628288B (en) | Performance evaluation method for first-order time-lag combined integral control system | |
| CN115938496A (en) | Quality Estimation Method Based on XGBoost Algorithm | |
| CN112859793B (en) | Industrial production process dynamic time delay identification method based on improved sliding time window | |
| Springer et al. | Robust parameter estimation of chaotic systems | |
| CN116842358A (en) | Soft measurement modeling method based on multi-scale convolution and self-adaptive feature fusion | |
| CN114580151A (en) | Water demand prediction method based on gray linear regression-Markov chain model | |
| CN116662925A (en) | Industrial process soft measurement method based on weighted sparse neural network | |
| CN116522065A (en) | Coal mill health degree assessment method based on deep learning | |
| CN113111588B (en) | NO of gas turbine X Emission concentration prediction method and device | |
| CN105069214A (en) | Process reliability evaluation method based on nonlinear correlation analysis | |
| CN114943328A (en) | SARIMA-GRU time sequence prediction model based on BP neural network nonlinear combination | |
| CN113347014B (en) | Industrial control system situation combination prediction method based on time sequence | |
| CN114021465A (en) | Electric power system robust state estimation method and system based on deep learning | |
| CN110673470B (en) | Industrial non-stationary process soft measurement modeling method based on local weighting factor model | |
| CN113849479A (en) | Comprehensive energy supply station oil tank leakage detection method based on instant learning and self-adaptive threshold | |
| CN111507003A (en) | Method for predicting key variables of debutanizer based on quality-related dynamic characteristic extraction | |
| CN118114812B (en) | Shale gas yield prediction method, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |