CN117520711A - Prediction method, prediction system and storage medium for melt index - Google Patents
Prediction method, prediction system and storage medium for melt index Download PDFInfo
- Publication number
- CN117520711A CN117520711A CN202311543417.8A CN202311543417A CN117520711A CN 117520711 A CN117520711 A CN 117520711A CN 202311543417 A CN202311543417 A CN 202311543417A CN 117520711 A CN117520711 A CN 117520711A
- Authority
- CN
- China
- Prior art keywords
- melt index
- model
- semi
- empirical formula
- loop
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 74
- 238000003860 storage Methods 0.000 title claims abstract description 9
- 238000012937 correction Methods 0.000 claims abstract description 114
- 229920000642 polymer Polymers 0.000 claims abstract description 98
- 238000005457 optimization Methods 0.000 claims abstract description 90
- 239000000155 melt Substances 0.000 claims abstract description 83
- 238000004519 manufacturing process Methods 0.000 claims abstract description 27
- 230000007246 mechanism Effects 0.000 claims abstract description 25
- 238000006116 polymerization reaction Methods 0.000 claims abstract description 21
- 230000001186 cumulative effect Effects 0.000 claims description 38
- 238000005259 measurement Methods 0.000 claims description 20
- 239000003054 catalyst Substances 0.000 claims description 18
- 230000004044 response Effects 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 9
- 230000002068 genetic effect Effects 0.000 claims description 8
- 238000010187 selection method Methods 0.000 claims description 8
- 238000009825 accumulation Methods 0.000 abstract 1
- 230000008569 process Effects 0.000 description 23
- 239000007789 gas Substances 0.000 description 10
- 238000010586 diagram Methods 0.000 description 7
- 238000005070 sampling Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 6
- 238000012549 training Methods 0.000 description 5
- 239000004743 Polypropylene Substances 0.000 description 4
- -1 polypropylene Polymers 0.000 description 4
- 229920001155 polypropylene Polymers 0.000 description 4
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000001257 hydrogen Substances 0.000 description 3
- 229910052739 hydrogen Inorganic materials 0.000 description 3
- QQONPFPTGQHPMA-UHFFFAOYSA-N propylene Natural products CC=C QQONPFPTGQHPMA-UHFFFAOYSA-N 0.000 description 3
- 125000004805 propylene group Chemical group [H]C([H])([H])C([H])([*:1])C([H])([H])[*:2] 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013499 data model Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000282326 Felis catus Species 0.000 description 1
- 238000003556 assay Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000007790 solid phase Substances 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- 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
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Educational Administration (AREA)
- Pure & Applied Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computational Mathematics (AREA)
- Algebra (AREA)
- Manufacturing & Machinery (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a melt index prediction method, a melt index prediction system and a computer readable storage medium. The method for predicting the melt index comprises the following steps: obtaining the instantaneous property of the polymer in each loop based on a polymerization reaction mechanism, and establishing a quantitative relation between the instantaneous property and the accumulation property of the polymer in each loop based on a mass conservation law of a multi-loop production process so as to obtain a semi-empirical formula model; establishing an SVR data correction model by utilizing an error between an output value of the semi-empirical formula model and an actual melt index measured value; optimizing the semi-empirical formula model to determine first optimization parameters of the semi-empirical formula model, and optimizing the SVR data correction model based on the first optimization parameters to obtain second optimization parameters of the SVR data correction model; and determining the melt index in real time according to the online measured value of the operating variable based on the optimized semi-empirical formula model and the SVR data correction model.
Description
Technical Field
The present invention relates to the field of polymer production, and more particularly, to a method for predicting melt index, a system for predicting melt index, and a computer-readable storage medium.
Background
Along with the diversification of the market demand for polymer products, petrochemical enterprises are required to have the capability of frequently and continuously switching different brands of products for production in the polymer production process so as to maximize the economic benefit of the enterprises.
When the polymerization reaction is carried out, the quality of the polymer in each reactor loop cannot be estimated in real time, so that various quality indexes of the polymer are obtained through manual sampling and off-line assay analysis in the prior art, and meanwhile, the residence time of the polymer in each reactor loop is long (generally 2-4 hours). This results in a long duration of the different brands of switching process, and the quality indicators of the polymer at the outlet of each reactor loop also require a period of time before the product requirements are met, and during this period a large number of off-grade transitions are produced in each reactor loop, resulting in a large economic loss.
Among the quality indexes, melt Index (MI) is one of the key indexes in quality monitoring, and the measurement of melt index is an important link for controlling the quality of products in the production process of polymers. Thus, the switching process for different brands in polymer production can be optimized by determining the melt index of the polymer at the outlet of each reactor loop in real time.
Disclosure of Invention
The following presents a simplified summary of one or more aspects in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects, and is intended to neither identify key or critical elements of all aspects nor delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.
In order to overcome the above-mentioned drawbacks of the prior art, the present invention provides a melt index prediction method, a melt index prediction system, and a computer-readable storage medium. The prediction technology of the melt index uses a semi-empirical formula model as a main model, an SVR data correction model as an auxiliary model, and then an optimization algorithm is used for finding out optimization parameters of the semi-empirical formula model and the SVR data correction model, so that the prediction method of the real-time predicted melt index has the advantages of small sample size required by a mechanism model and high precision of a data model, and improves the prediction precision while reducing the modeling difficulty.
Specifically, the method for predicting the melt index provided in accordance with the first aspect of the present invention includes the steps of: obtaining the instantaneous property of the polymer in each loop based on a polymerization reaction mechanism, and establishing a quantitative relation between the instantaneous property and the accumulated property of the polymer in each loop based on a mass conservation law of a multi-loop production process so as to obtain a semi-empirical formula model; determining a correction term through an error between an output value of the semi-empirical formula model and an actual melt index measurement value, and establishing an SVR data correction model by using the correction term; optimizing the semi-empirical formula model to determine first optimization parameters of the semi-empirical formula model, and optimizing the SVR data correction model based on the first optimization parameters to obtain second optimization parameters of the SVR data correction model; and determining the melt index in real time according to the online measured value of the operation variable based on the optimized semi-empirical formula model and the SVR data correction model.
Further, in an embodiment of the present invention, the obtaining the instantaneous property of the polymer in each loop based on the polymerization mechanism, and establishing the quantitative relationship between the instantaneous property and the cumulative property of the polymer in each loop based on the mass conservation law of the multi-loop production process to obtain the semi-empirical formula model further includes the steps of: determining an instantaneous melt index of the polymer in each loop based on the instantaneous properties; the cumulative melt index in each loop is determined based on a quantitative relationship of the instantaneous and cumulative properties of the polymer in each loop.
Further, in an embodiment of the present invention, obtaining the instantaneous properties of the polymer in each loop based on the polymerization mechanism, and establishing a quantitative relationship between the instantaneous properties and the cumulative properties of the polymer in each loop based on the mass conservation law of the multi-loop production process, to obtain a semi-empirical formula model, further comprises the steps of: the semi-empirical formula model is simplified by a finite impulse response model.
Further, in an embodiment of the present invention, the determining a correction term by an error between the output value of the semi-empirical formula model and the actual melt index measurement value, and the building the SVR data correction model using the correction term further includes the steps of: and selecting a radial basis function as a kernel function of the SVR data correction model.
Further, in an embodiment of the present invention, the loop includes a pre-polymerization loop, the determining a correction term by an error between the output value of the semi-empirical formula model and the actual melt index measurement value, and the building the SVR data correction model using the correction term further includes the steps of: and taking the catalyst consumption and the catalyst flow of the prepolymerization loop as input variables of the SVR data correction model.
Further, in an embodiment of the present invention, the optimizing the semi-empirical formula model to determine the first optimization parameters of the semi-empirical formula model further comprises the steps of: searching the minimum correction term through a first optimization problem of the semi-empirical formula model; solving a first optimization problem by using a genetic algorithm and a roulette selection method; and determining a first optimization parameter of the semi-empirical formula model using the first optimization problem.
Further, in an embodiment of the present invention, the optimizing the SVR data correction model based on the first optimization parameter to obtain the second optimization parameter of the SVR data correction model further includes the steps of: substituting the first optimization parameter into the semi-empirical formula model, and determining the optimized semi-empirical formula model and a first estimated value of the melt index; determining a second estimate of the melt index based on the first estimate of the melt index and the correction term; searching an error between a second estimated value of the minimum melt index and an actual melt index measured value through a second optimization problem of the SVR data correction model; solving a second optimization problem by using a genetic algorithm and a roulette selection method; and utilizing the second optimization problem to obtain a second optimization parameter of the SVR data correction model.
In addition, the above melt index prediction system provided according to the second aspect of the present invention includes a memory and a processor. The memory has stored thereon computer instructions. The processor is coupled to the memory and configured to execute computer instructions stored on the memory to implement the melt index prediction method provided in any one of the embodiments described above.
Further, the above-described computer-readable storage medium according to the third aspect of the present invention has stored thereon computer instructions. The computer instructions, when executed by a processor, implement the melt index prediction method provided in any one of the embodiments above.
Drawings
The above features and advantages of the present invention will be better understood after reading the detailed description of embodiments of the present disclosure in conjunction with the following drawings. In the drawings, the components are not necessarily to scale and components having similar related features or characteristics may have the same or similar reference numerals.
FIG. 1 illustrates a flow chart of a method of predicting melt index provided in accordance with some embodiments of the present invention;
FIG. 2 illustrates a schematic diagram of multiple ring pipe series during a polymer reaction provided in accordance with some embodiments of the present invention;
FIG. 3 illustrates a block diagram of a melt index prediction system provided in accordance with some embodiments of the present invention;
FIG. 4 illustrates an optimization flow chart of a melt index prediction method provided in accordance with some embodiments of the invention;
FIG. 5 illustrates a schematic diagram of the results of a method of predicting melt index provided in accordance with some embodiments of the present invention; and
fig. 6 illustrates a schematic diagram of the results of a method of predicting melt index provided in accordance with some embodiments of the present invention.
Detailed Description
The invention is described in detail below with reference to the drawings and the specific embodiments. It is noted that the aspects described below in connection with the drawings and the specific embodiments are merely exemplary and should not be construed as limiting the scope of the invention in any way.
In the description of the present invention, it should be noted that, unless explicitly specified and limited otherwise, the terms "mounted," "connected," and "connected" are to be construed broadly, and may be either fixedly connected, detachably connected, or integrally connected, for example; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
In addition, the terms "upper", "lower", "left", "right", "top", "bottom", "horizontal", "vertical" as used in the following description should be understood as referring to the orientation depicted in this paragraph and the associated drawings. This relative terminology is for convenience only and is not intended to be limiting of the invention as it is described in terms of the apparatus being manufactured or operated in a particular orientation.
It will be understood that, although the terms "first," "second," "third," etc. may be used herein to describe various elements, regions, layers and/or sections, these elements, regions, layers and/or sections should not be limited by these terms and these terms are merely used to distinguish between different elements, regions, layers and/or sections. Accordingly, a first component, region, layer, and/or section discussed below could be termed a second component, region, layer, and/or section without departing from some embodiments of the present invention.
As mentioned above, by determining the melt index of the polymer at the outlet of each reactor loop in real time, it can be used to optimize the switching process for different brands in the production of the polymer. Accordingly, the present invention provides a method of predicting a melt index, a system of predicting a melt index, and a computer-readable storage medium. The prediction technology of the melt index uses a semi-empirical formula model as a main model, an SVR data correction model as an auxiliary model, and then an optimization algorithm is used for finding out optimization parameters of the semi-empirical formula model and the SVR data correction model, so that the prediction method of the real-time predicted melt index has the advantages of small sample size required by a mechanism model and high precision of a data model, and improves the prediction precision while reducing the modeling difficulty.
In some non-limiting embodiments, the above-described melt index prediction method provided by the first aspect of the present invention may be implemented via the above-described melt index prediction system provided by the second aspect of the present invention. Specifically, the melt index prediction system may be provided with a memory and a processor. The memory includes, but is not limited to, the above-described computer-readable storage medium provided by the third aspect of the present invention, having stored thereon computer instructions. The processor is coupled to the memory and configured to execute computer instructions stored on the memory to implement the melt index prediction method provided in the first aspect of the present invention.
The principle of operation of the above melt index prediction system will be described below first in connection with some embodiments of the melt index prediction method. It will be appreciated by those skilled in the art that these examples of melt index prediction methods are merely some non-limiting embodiments provided by the present invention, and are intended to clearly illustrate the general concepts of the present invention and to provide some embodiments for public implementation, not to limit the overall functionality or overall operation of the melt index prediction system. Similarly, the melt index prediction system is just one non-limiting embodiment provided by the present invention, and does not limit the execution subjects and execution orders of the steps in the melt index prediction methods.
Referring to fig. 1, fig. 1 is a flowchart illustrating a method for predicting a melt index according to some embodiments of the present invention.
As shown in fig. 1, the method 100 for predicting the melt index shown in fig. 1 includes the steps of:
s110: the instantaneous property of the polymer in each loop is obtained based on a polymerization reaction mechanism, and a quantitative relation between the instantaneous property and the accumulated property of the polymer in each loop is established based on the mass conservation law of the multi-loop production process, so that a semi-empirical formula model is obtained.
In one non-limiting example, the data during the production of the polymer includes process operating variables such as catalyst, pressure, temperature, and gas concentration of each loop, and index variables such as actual measurements of the melt index of the polymer in the final loop. The individual operating variables can be read via a control device such as a distributed control system (DCS, distributed Control System), and the time interval for the individual operating variables is typically several seconds, but the time interval for the cumulative melt index is 2 to 4 hours. The melt index prediction system may select an operating variable for predicting the melt index from data read by the control device based on various factors affecting the melt index during the polymerization reaction.
The instantaneous properties of the polymer in each loop are then obtained based on the polymerization mechanism. Referring specifically to fig. 2, fig. 2 illustrates a schematic diagram of multiple loop pipes in series during a polymer reaction process according to some embodiments of the present invention.
As shown in FIG. 2, in one non-limiting example, the polymer may be produced using i+1 loops in series, each loop may be fed with n gases through MIXERs (MIXER) M (1) to M (i+1) to effect polymerization. Here, the melt index prediction system may select the temperature of each loop and the different component gases in each loop in the actual production of the polymer as input variables to determine the instantaneous properties of the polymer in each loop, where the input variables are shown in table 1.
Table 1 input variables required to calculate transient properties
Based on the polymerization mechanism, the melt index prediction system may select a model of instantaneous properties with linear characteristics, and combine with the input variables to determine the quantitative relationship between instantaneous melt index and gas concentration of the polymer in the j-th loop:
wherein t is time, MI i,j (t) is an estimate of the instantaneous melt index, K, of the polymer in the jth loop at time t p j (p=0, 1,2, …, n) is a parameter to be estimated for the instantaneous properties of the polymer in the j-th loop,for the jth loop, the nth gas concentration, T is the jth loop temperature.
After the instantaneous properties of the polymer in each loop are obtained, the quantitative relation between the instantaneous properties and the accumulated properties of the polymer in each loop is established based on the mass conservation law of the multi-loop production process. In the embodiment shown in FIG. 2, for the jth loop, first, the accumulated polymer formed in the jth-1 loop continuously enters the jth loop at a certain flow rate. Second, the jth loop itself also generates new polymer in real time. Assuming that the weight of polymer contained in the jth loop does not change during a very small period of time, the sum of the flow rate of polymer at the inlet of the jth loop and the rate of formation of new polymer is equal to the flow rate at the outlet of the jth loop. Further, after the time period h, the polymer contained in the jth loop is composed of the newly generated polymer in the time period h, the polymer at the j-1 th loop outlet and the polymer originally contained in the jth loop.
Thus, the inlet polymer flow of one loop, i.e. the outlet flow of the last loop polymer, is added to the amount of new polymer produced by the loop itself, i.e. equal to the outlet flow of the loop polymer. From this relationship, a quantitative relationship of instantaneous properties to cumulative properties can be obtained. Wherein the calculation of the melt index of the polymer in the individual loops is given by the differential equation:
wherein,M wc is the cumulative average molecular weight, M, of the loop polymer wi Is the instantaneous average molecular weight of the loop polymer, τ is the solid phase residence time, MI is the melt index of the loop polymer, M w Is the average molecular weight, MI and M of the loop polymer w Inversely proportional, the index is about 3.5.
In the embodiment shown in fig. 2, the melt index prediction system can determine the estimated cumulative melt index of the polymer in loop 1 and loop j from this map and solving the differential equation above:
wherein t is time, h is the length of time traced back from a certain melt index measurement point, MI c,1 MI is an estimate of the cumulative melt index of the 1 st loop c,j MI is an estimate of the cumulative melt index of the jth loop i,j Is an estimate of the instantaneous melt index, τ, of the polymer in the j-th loop j-1 、τ j The residence time distribution of the j-1 th and j-th loops, respectively.
Further, a semi-empirical formula model is obtained from the instantaneous properties and quantitative relationships of the instantaneous properties to the cumulative properties, and a mechanism model of the relationship of the on-line measurement of the manipulated variables to the cumulative melt index of the polymer at the outlet of the last reactor loop can be established based thereon. The mechanism model belongs to a typical autoregressive model, and the dynamic characteristics of the polymerization process are characterized by a polymerization mechanism. The production operating conditions of different brands are different, but all follow the basic principle of polymerization. Thus, a mechanism model can predict a plurality of different brands of production operation conditions, especially in the dynamic process of brand switching.
In a preferred embodiment, the melt index prediction system can simplify the complex autoregressive recurrence relation into a finite impulse response model (FIR, finite Impulse Response) and the semi-empirical formula model by the finite impulse response model, since the autoregressive model is prone to model instability when calculated in a computer.
Specifically, the finite impulse response model may be set to:
wherein T ', K and tau are parameters of a semi-empirical formula model, T' is a time constant, K is a proportionality coefficient, tau is a pure time delay constant, and L is a length parameter.
Further, the quantitative relationship between the instantaneous and cumulative properties of the polymer in each loop is expressed by the following expression, and the estimated value of the cumulative melt index in the 1 st and j th loops can be determined therefrom:
wherein MI is c,1 (mq) (m=1, …, N) is an estimate of the cumulative melt index of the polymer in the first loop at the time of mq, MI c,j (mq) (j=2, …, M) is an estimate of the cumulative melt index of the polymer in the jth loop at the time of mq; MI (MI) i,1 (mq-s) is an estimate of the instantaneous melt index, MI, of the polymer in the 1 st loop at the time of mq-s i,j (mq-s) is an estimate of the instantaneous melt index of the polymer in the jth loop at the time of the mq-s, q is the sampling period (typically 2-4 hours) of the cumulative melt index of the polymer in the last loop, α is a weight coefficient for measuring the ratio of the outlet flow of the last loop polymer to the outlet flow of the current loop polymer, h(s) is a finite impulse response parameter, and L is a length parameter.
Wherein the instantaneous melt index and the cumulative melt index samples make up a time series of samples as follows:
the temporal sequence of instantaneous melt index samples is expressed as:
[MI i,j (1),MI i,j (2),...,MI i,j (q),MI i,j (q+1),...,MI i,j (2q),...,MI i,j (mq)]
the cumulative melt index sampling time series is expressed as:
[MI c,j (q),MI c,j (2q),...,MI c,j (mq)]
the mechanism modeling involves a large number of complex differential equations, and meanwhile, the accuracy requirements on physicochemical parameters are high. Simplifying the autoregressive model into a finite impulse response model can avoid the integration process in the mechanism model. For the cumulative nature of adjacent loops, the use of weighting coefficients for the finite impulse response model to replace the original flow relationship may eliminate the need for more prior knowledge in establishing the model.
With continued reference to fig. 1, the method 100 for predicting the melt index shown in fig. 1 includes the steps of:
s120: and determining a correction term through an error between an output value of the semi-empirical formula model and an actual melt index measured value, and establishing an SVR data correction model by using the correction term.
The prediction system of the melt index determines an error between an output value of the semi-empirical formula model and an actual melt index measurement value as a correction term, and uses the correction term as an output variable of the SVR data correction model to establish the SVR data correction model. In a preferred embodiment, the melt index prediction system may take a correction term of the error between the estimated value of the accumulated melt index of the last loop in the semi-empirical formula model and the actually measured accumulated melt index as the output variable of the SVR data correction model.
In addition, when the operation variable is read, the prediction system of the melt index has a longer interval of time for accumulating the melt index and a larger interval of time for taking the melt index from a common value. The melt index prediction system combines the data of all the process operating variables such as the catalyst, the pressure, the temperature, the gas concentration and the like of each loop in a period of time before each accumulated melt index measurement point, and uses the data as the input vector of the second-stage SVR data correction model.
With continued reference to FIG. 2, in the embodiment shown in FIG. 2, distinguishing between the key operating variables for each brand is a catalyst related factor. Because the semi-empirical formula model does not consider key operating variables such as catalyst factors and the like for distinguishing each brand, the mechanism model constructed by the semi-empirical formula model can only predict the melt index of a single brand. In this embodiment, after the catalyst factors are introduced into the SVR data correction model, the melt index prediction system may further implement prediction of melt indexes of different brands with different catalyst contents in the same catalyst system. Thus, the melt index prediction system may take the catalyst usage and catalyst flow of the prepolymerization loop as input variables to the SVR data correction model to correct for structural deviations in the semi-empirical formula model.
In a preferred embodiment, the prediction system of the melt index may choose a radial basis function as the kernel function of the SVR data correction model, taking into account the strong nonlinear relationship between the input variables and the output variables of the SVR data correction model.
In the embodiment shown in FIG. 2, the melt index prediction system selects the radial basis function as the kernel function of the SVR data correction model to determine the concentration of each gas component in each loopCatalyst amount N cat Amount of cocatalyst N donor Temperature T inside the jth loop j And internal pressure P j As an input variable to the SVR data correction model, and determining a correction term for the error between the estimated value of the cumulative melt index of the polymer in the last loop in the semi-empirical formula model and the actually measured cumulative melt index +.>To correct item->As an output variable of the SVR data correction model.
As such, the SVR data correction model may be expressed as:
wherein,for the input variable of the SVR data correction model, h is the time length traced back from a certain melt index measurement point, c is the regularization coefficient of the SVR data correction model, and gamma is the coefficient of the radial basis function.
In the preferred embodiment, the SVR data corrects the input variables of the modelThe time series of input variables may correspond to a cumulative melt index sampling time series based on all measurements of the selected operating variable over a period of time. Here, the time-series expression of the input variable may be as follows:
the time-series expression of the output variable may be as follows:
wherein,the error between the estimated and actual value of the cumulative melt index for the last loop exit polymer.
After the semi-empirical formula model and the SVR data correction model are established, a large number of parameters to be estimated exist in the two models. With continued reference to fig. 1, the method 100 for predicting the melt index shown in fig. 1 includes the steps of:
s130: and optimizing the semi-empirical formula model to determine first optimization parameters of the semi-empirical formula model, and optimizing the SVR data correction model based on the first optimization parameters to obtain second optimization parameters of the SVR data correction model.
Referring to fig. 3, fig. 3 illustrates an architecture diagram of a melt index prediction system provided in accordance with some embodiments of the present invention.
In the embodiment shown in fig. 3, the overall model of the melt index prediction system is composed of a semi-empirical formula model that plays a dominant role in the prediction process and a SVR data correction model that plays an auxiliary role in the prediction process. In the preferred embodiment, the semi-empirical formula model can be reduced to a finite impulse response model (FIR).
In a reference example, quality index is generally difficult to obtain in practical process in polymer production process such as polypropylene due to lack of effective on-line measurement method, and sampling frequency of index variable in production process is far lower than that of operation variable, so that a large number of unlabeled samples are generated, and difficulty is brought to melt index prediction of brand switching dynamic process.
Unlike mechanism modeling, the SVR data correction model is built without knowledge of the polymer reaction mechanism, and only relies on process data, but the data correction model requires a large amount of sample. And, since the SVR data correction model ignores the polymer reaction mechanism, the interpretation of the SVR data correction model constructed alone is poor. The polymerization reaction has the characteristics of strong nonlinearity and large uncertainty, and the production time of a single brand in actual production is short, and the number of effective samples is small. Simple data-driven methods have difficulty in rapidly and accurately predicting the melt index of a polymer in such a case. Therefore, the SVR data correction model is taken as an auxiliary model, and the accuracy of the semi-empirical formula model playing a leading role can be improved under the condition of small sample size, so that the prediction method of the melt index can have the advantages of small sample size required by the semi-empirical formula model and high accuracy of the SVR data correction model.
In order to ensure the dominant effect of the semi-empirical formula model in the prediction process, parameter optimization is respectively carried out in two stages in the process of iterative updating of the model parameters of a finite impulse response model (FIR) and SVR. First, a semi-empirical formula model is optimized to determine a first optimization parameter of the semi-empirical formula model. And optimizing the SVR data correction model based on the first optimization parameters of the semi-empirical formula model to obtain second optimization parameters of the SVR data correction model. And finally, determining the optimization parameters of the total model through the first optimization parameters and the second optimization parameters.
Referring to fig. 4, fig. 4 illustrates an optimization flow chart of a melt index prediction method provided in accordance with some embodiments of the present invention.
The prediction system of melt index may first optimize the semi-empirical formula model to determine a first optimization parameter of the semi-empirical formula model. Specifically, in the embodiment shown in FIG. 4, the melt index prediction system may first generate an initial population of parameter values from the manipulated variables and based on a semi-empirical formula model, and calculate fitness values for each individual using a genetic algorithm. And selecting individuals with large fitness from the population to enter the next generation population by using a roulette selection method, and performing cross operation and mutation operation on the next generation population to generate a child population. Thereafter, fitness values of individuals of the offspring population are calculated. And repeatedly and continuously iterating the process until the optimal parameter value of the semi-empirical formula model is found after the stopping criterion is met.
The melt index prediction system may further optimize the semi-empirical formula model by determining a specific first optimization problem and utilizing the first optimization problem to determine a first optimization parameter of the semi-empirical formula model. In an alternative embodiment, the first optimization problem may be as follows:
wherein,polymerization of the last loop for a semi-empirical formula modelEstimate of the cumulative melt index, MI, of the article at the mq-th moment c (mq) is the actual measurement of the cumulative melt index of the polymer of the last loop at the time of the mq, -, I->Is an estimate of the instantaneous melt index of the polymer of the penultimate loop at the time mq-s,/I>Is an estimate of the cumulative melt index of the polymer of the penultimate loop at the time mq-s,/I>Is the predicted output value of the accumulated melt index of the polymer of the j-th loop at the mq time,predicted output value of instantaneous melt index of polymer of jth loop at the Mq time,/>For the concentration of the n-th gas in the loop, h(s) is the impulse response parameter, T', ->τ and K are first optimization parameters to be optimized for the first optimization problem.
And searching the smallest correction term through the first optimization problem of the semi-empirical formula model, and further solving the first optimization problem by using a genetic algorithm and a roulette selection method to determine the first optimization parameter of the semi-empirical formula model.
Please continue to refer to fig. 4, after obtaining the optimal parameters of the semi-empirical formula model, substituting the optimal parameters into the SVR data correction model, and generating an initial population of parameter values based on the SVR data correction model. In a preferred embodiment, the melt index prediction system may solve the first optimization problem described aboveOptimizing the obtained T'τ, K are substituted into the SVR data correction model.
And then, calculating the fitness value of each individual by using a genetic algorithm. And selecting individuals with large fitness from the population to enter the next generation population by using a roulette selection method, and performing cross operation and mutation operation on the next generation population to generate a child population. Thereafter, fitness values of individuals of the offspring population are calculated. And repeatedly and continuously iterating the process until the optimal parameter value of the SVR data correction model is found after the stopping criterion is met.
The melt index prediction system may further optimize the SVR data correction model by determining a specific second optimization problem and utilizing the second optimization problem to determine a second optimization parameter of the SVR data correction model.
Specifically, the first estimated value of the optimization latter half empirical formula model and the correction term of the SVR data correction model constitute the second estimated value of the cumulative melt index at each time. Because the first estimated value of the optimized latter half empirical formula model is quite close to the real sampling value of the accumulated melt index, the correction term of the SVR data correction model after substituting the first optimization parameter, namely the actual value of the output variable of the SVR data correction model, is relatively smaller, so that the dominant position of the half empirical formula model in the total model in the prediction process is ensured.
There is still an error between the second estimate of the accumulated melt index at each instant and the true actual melt index measurement, which can be targeted by the melt index prediction system for the second optimization problem.
Here, the second optimization problem may be as follows:
wherein MI is c (mq) is a true measure of the cumulative melt index of the last loop polymer at time mq,for the first optimization parameter (e.g. T', +_>τ, K) are substituted into the estimated value of the cumulative melt index outputted after the semi-empirical formula model, +.>Correcting the correction term of the model for the SVR data of the last loop at the mq time,and predicting output values for the accumulated melt index at the mq time, wherein c and gamma are second optimization parameters for optimizing the second optimization problem.
And searching for an error between the second estimated value of the minimum melt index and the actual melt index measured value through the second optimization problem of the SVR data correction model, and further solving the second optimization problem by using a genetic algorithm and a roulette selection method to obtain a second optimization parameter of the SVR data correction model.
In summary, in a preferred embodiment, the first optimization parameters are obtained by optimizing the semi-empirical formula model with the first optimization problem, and the second optimization parameters are substituted into the SVR data correction model to obtain the second optimization parameters of the SVR data correction model, and finally all the optimization parameters of the total model are obtained, as shown in table 2.
Table 2 all optimization parameters of the total model
With continued reference to fig. 1, the method 100 for predicting the melt index shown in fig. 1 includes the steps of:
s140: and determining the melt index in real time according to the online measured value of the operating variable based on the optimized semi-empirical formula model and the SVR data correction model.
Specifically, an optimized semi-empirical formula model and an SVR data correction model are determined according to the first optimization parameter and the second optimization parameter. And then, determining the melt index in real time according to the online measured value of the operation variable based on the optimized semi-empirical formula model and the SVR data correction model. The melt index determined in real time is written into a database for reference by an operator to guide the actual operation or for controlling and optimizing the production flow of the polymer.
The method for predicting the melt index constructs the basic structure of the total model through the semi-empirical formula model and the SVR data correction model, and then determines the optimization parameters of the semi-empirical formula model and the SVR data correction model through an optimization algorithm to obtain the optimized semi-empirical formula model and SVR data correction model. Compared with the prior art, the method for predicting the melt index reduces modeling difficulty and improves prediction accuracy.
The following is a specific, non-limiting, preferred embodiment, according to which the method and system for predicting melt index of the present invention are further described.
Taking the polypropylene production process as an example, real factory data of an operation variable is obtained from a certain polypropylene production enterprise, and an online soft measurement model of the polypropylene melt index is established based on the data. In particular, the data used are actual production process data for the business for 5 brands of a certain homopolymerization process, the 5 brands of catalyst systems being identical but differing in catalyst usage. The homopolymerization process employs two loop reactors in series to produce a polymer product, and the gas components inside the two loops contain only hydrogen and propylene.
And selecting the temperatures of the first loop and the second loop and the hydrogen concentration and the propylene concentration of the two loops from the operation variables as the operation variables for predicting the melt index, and simultaneously recording the actual analysis values of the melt index corresponding to the operation variables, wherein the total of 1191 pieces of effective data is obtained.
Thereafter, the instantaneous properties of the polymer in each loop were determined, where the quantitative relationship of the instantaneous melt index of the two loop polymers to the gas concentration was:
according to the quantitative relation between the instantaneous properties and the accumulated properties of the polymers in the multiple loops, the instantaneous properties of the polymers in the first loop and the second loop and the accumulated properties of the polymers in the first loop can be utilized by the following formula through a simplified finite pulse corresponding model:
since there are only two loops, the cumulative melt index of the polymer in the second loop may be calculated by only relying on the cumulative melt index of the polymer in the first loop at the current time.
A time series of samples is formed by the obtained instantaneous melt index and the accumulated melt index samples of the dynamic process. Here, the temporal sequence of instantaneous melt index samples is expressed as:
the cumulative melt index sampling time series is expressed as:
then, the catalyst flow rate, the cocatalyst flow rate, the temperature of the first loop and the second loop and the concentration of hydrogen and propylene of the two loops are selected from the operation variables as the input variables of the SVR data correction modelThe amount, further, is to combine the data of all process operating variables for a period of time before each cumulative melt index measurement point as SVR data correction model input variables, i.eWherein,
determining correction term by error between output value of semi-empirical formula model and actual melt index measurement valueAnd as an output variable of the SVR data correction model.
And setting all parameter ranges of the total model, optimizing the semi-empirical formula model to determine first optimization parameters of the semi-empirical formula model, and optimizing the SVR data correction model based on the first optimization parameters to obtain second optimization parameters of the SVR data correction model, so that the optimized semi-empirical formula model and the SVR data correction model are determined through the first optimization parameters and the second optimization parameters.
Here, the first and second optimization problems of the optimization semi-empirical formula model and the SVR data correction model may be:
and based on the established model, carrying out accumulated melt index prediction on the test set. Referring to fig. 5 and 6, fig. 5 and 6 are schematic diagrams illustrating the result of the method for predicting the melt index according to some embodiments of the present invention.
As shown in fig. 5 and 6, the abscissa of the graph represents the observed data points, and the ordinate represents the Melt Index (MI) value. The solid line data are actual MI values, and the dotted line is a mixed model melt index MI predicted value of the melt index prediction method provided by the invention. In this embodiment, the number of training sets provided in fig. 5 is 800, and the test set is checked based on the method for predicting the melt index provided in the present invention, so that it can be found that the prediction effect of the method for predicting the melt index provided in the present invention is better when the number of training sets is sufficient.
Then, only one group of data is selected as a training set for each brand, and the number of the training sets is 53 samples. The whole data set is predicted again according to the situation when the number of training sets is small, as shown in fig. 6. The prediction result shows that the method for predicting the melt index provided by the invention can obtain an acceptable model with better precision by only needing a small amount of industrial data.
In summary, the method for predicting the melt index fully considers the structural information of the polymer reaction process, combines the SVR data correction model with the semi-empirical formula model, and combines the advantages of data modeling and mechanism modeling, so that the prediction result considers the mechanism characteristic of the polymer reaction process and can adapt to the data with larger difference of melt index values in the brand switching process, thereby effectively coping with uncertainty. The key operation variables of distinguishing marks such as catalyst factors which are not reflected by the mechanism model are introduced into the semi-empirical formula model through a data correction method, so that the method is better suitable for online measurement of quality indexes in the polymerization reaction process.
While, for purposes of simplicity of explanation, the methodologies are shown and described as a series of acts, it is to be understood and appreciated that the methodologies are not limited by the order of acts, as some acts may, in accordance with one or more embodiments, occur in different orders and/or concurrently with other acts from that shown and described herein or not shown and described herein, as would be understood and appreciated by those skilled in the art.
The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (9)
1. A method for predicting melt index, comprising the steps of:
obtaining the instantaneous property of the polymer in each loop based on a polymerization reaction mechanism, and establishing a quantitative relation between the instantaneous property and the accumulated property of the polymer in each loop based on a mass conservation law of a multi-loop production process so as to obtain a semi-empirical formula model;
determining a correction term through an error between an output value of the semi-empirical formula model and an actual melt index measurement value, and establishing an SVR data correction model by using the correction term;
optimizing the semi-empirical formula model to determine first optimization parameters of the semi-empirical formula model, and optimizing the SVR data correction model based on the first optimization parameters to obtain second optimization parameters of the SVR data correction model; and
and determining the melt index in real time according to the online measured value of the operation variable based on the optimized semi-empirical formula model and the SVR data correction model.
2. The method of predicting according to claim 1, wherein said deriving instantaneous properties of the polymer in each loop based on a polymerization mechanism and establishing quantitative relationships of said instantaneous properties and cumulative properties of said polymer in each loop based on a law of conservation of mass in a multi-loop production process to derive a semi-empirical formula model further comprises the steps of:
determining an instantaneous melt index of the polymer in each loop based on the instantaneous properties;
determining the cumulative melt index of the polymer in each loop based on a quantitative relationship of the instantaneous property to the cumulative property of the polymer in each loop.
3. The method of predicting according to claim 2, wherein deriving the instantaneous properties of the polymer in each loop based on a polymerization mechanism and establishing a quantitative relationship of said instantaneous properties and cumulative properties of the polymer in each loop based on the law of conservation of mass of the multi-loop production process to derive a semi-empirical formula model further comprises the steps of:
the semi-empirical formula model is simplified by a finite impulse response model.
4. The prediction method according to claim 1, wherein the determining a correction term by an error between the output value of the semi-empirical formula model and an actual melt index measurement value, and the building of the SVR data correction model using the correction term further comprises the steps of:
and selecting a radial basis function as a kernel function of the SVR data correction model.
5. The prediction method according to claim 1, wherein the loop comprises a pre-polymerization loop, wherein the determining a correction term by an error between an output value of the semi-empirical formula model and an actual melt index measurement value, and wherein the building of the SVR data correction model using the correction term further comprises the steps of:
and taking the catalyst consumption and the catalyst flow of the prepolymerization loop as input variables of the SVR data correction model.
6. The method of predicting according to claim 1, wherein said optimizing said semi-empirical formula model to determine a first optimization parameter of said semi-empirical formula model further comprises the steps of:
searching the minimum correction term through a first optimization problem of the semi-empirical formula model;
solving a first optimization problem by using a genetic algorithm and a roulette selection method; and
and determining a first optimization parameter of the semi-empirical formula model by using the first optimization problem.
7. The prediction method of claim 6, wherein said optimizing said SVR data correction model based on said first optimization parameters to obtain second optimization parameters for said SVR data correction model further comprises the steps of:
substituting the first optimization parameter into the semi-empirical formula model, and determining the optimized semi-empirical formula model and a first estimated value of the melt index;
determining a second estimate of the melt index based on the first estimate of the melt index and the correction term;
searching an error between a second estimated value of the minimum melt index and an actual melt index measured value through a second optimization problem of the SVR data correction model;
solving a second optimization problem by using a genetic algorithm and a roulette selection method; and
and utilizing the second optimization problem to obtain a second optimization parameter of the SVR data correction model.
8. A melt index prediction system, comprising:
a memory having stored thereon computer instructions; and
a processor connected to the memory and configured to execute computer instructions stored on the memory to implement the melt index prediction method of any one of claims 1 to 7.
9. A computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the method of predicting melt index according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311543417.8A CN117520711A (en) | 2023-11-17 | 2023-11-17 | Prediction method, prediction system and storage medium for melt index |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311543417.8A CN117520711A (en) | 2023-11-17 | 2023-11-17 | Prediction method, prediction system and storage medium for melt index |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117520711A true CN117520711A (en) | 2024-02-06 |
Family
ID=89752740
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311543417.8A Pending CN117520711A (en) | 2023-11-17 | 2023-11-17 | Prediction method, prediction system and storage medium for melt index |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117520711A (en) |
-
2023
- 2023-11-17 CN CN202311543417.8A patent/CN117520711A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sharmin et al. | Inferential sensors for estimation of polymer quality parameters: Industrial application of a PLS-based soft sensor for a LDPE plant | |
| CN107908891B (en) | A kind of Hydrological Time Series rejecting outliers method based on ARIMA-SVR | |
| Tham et al. | Soft-sensors for process estimation and inferential control | |
| Davidson et al. | Interpreting shock tube ignition data | |
| US7058617B1 (en) | Method and apparatus for training a system model with gain constraints | |
| MXPA02012834A (en) | Multi-variable matrix process control. | |
| CN101957598A (en) | Gray model-free control method for large time lag system | |
| CN106897775A (en) | Soft-measuring modeling method based on Bayes's integrated study | |
| Nakaya et al. | On-line tracking simulator with a hybrid of physical and Just-In-Time models | |
| CN110957011B (en) | Online production parameter estimation method of continuous stirring reactor under unknown time-varying measurement noise | |
| Hocalar et al. | Comparison of different estimation techniques for biomass concentration in large scale yeast fermentation | |
| CN100511042C (en) | Groove type reactor non-linear predication control method based on multi-kernel support vector machine | |
| CN111142494B (en) | Intelligent control method and system for amine liquid regeneration device | |
| CN107085371A (en) | Crude(oil)unit economic model forecast control method based on data-driven | |
| CN104408317A (en) | Metallurgy enterprise gas flow interval predicting method based on Bootstrap echo state network integration | |
| Duarte et al. | Combined mechanistic and empirical modelling | |
| CN109858707B (en) | ABC-ANFIS-CTF-based intelligent coking diagnosis method for furnace tube of ethylene cracking furnace | |
| CN101673096A (en) | Soft-measuring method for density in concentration process of salvia miltiorrhiza injection production | |
| Song et al. | Identification of multiple operating modes based on fused features for continuous annealing processes | |
| CN117520711A (en) | Prediction method, prediction system and storage medium for melt index | |
| Alvarez et al. | Batch process monitoring in the original measurement's space | |
| CN109033524A (en) | A kind of chemical process concentration variable On-line Estimation method based on robust mixed model | |
| Cisneros-Garibay et al. | Accounting for uncertainty in RCCE species selection | |
| CN111650894A (en) | Bayesian network complex industrial process soft measurement method based on hidden variables | |
| Valappil et al. | A systematic tuning approach for the use of extended Kalman filters in batch processes |
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 |