Chemical process yield prediction simulation method that considers lifespan factors

Abstract

The objective of the present invention is to provide a yield prediction simulation system and method that can predict the yield of a second cycle period on the basis of process operation data of a first cycle period and simulate yield prediction by reflecting tag variation. In addition, the objective of the present invention is to provide a yield prediction simulation system and method that divide a first cycle period into a plurality of segments on the basis of a catalyst lifespan and a yield change resulting therefrom, analyze key factors, predict yield and analyze tag variation for each segment, and simulate yield prediction for the second cycle period according thereto, and thus can further improve prediction accuracy over an entire cycle period.

Description

A simulation method for predicting yields in chemical processes considering life factors

[0001] The present invention relates to yield prediction of a chemical process, and more specifically, to a yield prediction simulation method capable of predicting the yield of a second cycle period based on process operation data of a first cycle period and simulating the yield prediction by reflecting tag fluctuations.

[0002] When performing a series of processes consisting of multiple steps, ensuring the integrity and reliability of each process through organic connections is crucial. To achieve this integrity, the process requires the development of an efficient process management system capable of identifying input values ​​from key equipment, identifying abnormalities in each process, and diagnosing their root causes.

[0003] Historical data generated in industrial processes typically consisted of a small number of variables and had linear data structures, making it possible to obtain sufficient prediction and classification results using existing algorithms. However, with the advancement of ICT and sensor technology, data with hundreds or thousands of variables are now being generated in the manufacturing process field. In particular, modern industrial processes, such as chemical and manufacturing processes and power plants, are increasingly growing in scale and complexity due to various efforts to reduce costs and maximize profits while simultaneously meeting safety, health, and environmental regulations.

[0004] Therefore, among this complex and massive volume of data, it is crucial to select and manage only the data that significantly impacts profit generation. For example, in commercial chemical processes, process operating conditions affect catalyst activity in both the short and long term, and catalyst activity is directly linked to product yield. Therefore, predicting catalyst activity from both a short- and long-term perspective is crucial.

[0005] From a short-term perspective, it's crucial to understand changes in catalyst activity depending on process operating conditions and optimize operating conditions to improve catalyst activity and increase product production. From a long-term perspective, in commercial chemical processes using catalysts, catalysts deactivate over time and lose activity, necessitating replacement after a certain period of time. Because catalyst replacement is time-consuming and costly, predicting future catalyst activity and determining catalyst lifespan / replacement timing is crucial.

[0006] Therefore, there is a need to develop new technologies that can identify process operating conditions (key factors) that significantly affect catalytic activity and reflect these conditions to improve the accuracy of predicting catalytic reaction activity.

[0007] (Patent Document 1) Patent Document 1: Korean Patent Publication No. 10-2018-0029114 (Published on March 20, 2018)

[0008] (Patent Document 2) Patent Document 2: Korean Patent No. 10-2222125 (announced on March 3, 2021)

[0009] (Patent Document 3) Patent Document 3: Korean Patent Publication No. 10-2018-0061769 (Published on June 8, 2018)

[0010] According to the present invention, it is an object to provide a yield prediction simulation system and method capable of predicting the yield of a second cycle period based on process operation data of a first cycle period and simulating the yield prediction by reflecting tag fluctuations.

[0011] According to the present invention, the first cycle period is divided into a plurality of segments based on the catalyst life and the resulting yield change, and key factor analysis, yield prediction, and tag variation analysis are performed for each segment, and a yield prediction simulation for the second cycle period is performed accordingly, thereby providing a yield prediction simulation system and method capable of further improving the prediction accuracy over the entire cycle period.

[0012] Other purposes of the present invention are not limited to the purposes mentioned above, and other purposes not mentioned will be clearly understood by those skilled in the art from the description below.

[0013] According to one embodiment of the present invention, a yield prediction simulation method for predicting a yield of a second cycle based on yield-related data of a first cycle in a chemical process is disclosed, the method comprising: a step of predicting the yield of the second cycle using a machine-learned yield prediction model using data of the first cycle as learning data; a tag variation analysis step of calculating a variation in at least some of the remaining input variables due to a variation in one or more of the input variables input to the yield prediction model as an operating condition of the chemical process; and a step of calculating a simulated yield of the second cycle based on the predicted yield calculated in the yield prediction step and the tag variation analysis result calculated in the tag variation analysis step.

[0014] In one embodiment, the yield prediction simulation method may further include a life factor analysis step for calculating a life factor reflecting the life of a catalyst used in a chemical process for at least a portion of a second cycle, wherein the simulation yield calculation step calculates a simulation yield based on the predicted yield, the tag variation analysis result, and the life factor for at least a portion of the second cycle.

[0015] In one embodiment, the simulation yield calculation step may calculate a final simulation yield by multiplying the simulation yield calculated based on the predicted yield and the tag variation analysis result by a weight calculated from the life factor for at least a portion of the second cycle.

[0016] In one embodiment, the life factor can be calculated by a formula including (daily propylene production) / (heat applied to catalyst).

[0017] In addition, in one embodiment, the life factor analysis step may include: calculating a life factor based on the amount of production compared to the amount of heat applied to the catalyst; dividing the calculated life factor into predetermined time intervals; and normalizing the average life factor within each interval; and scaling the normalized life factor.

[0018] According to one embodiment of the present invention, a computer-readable recording medium having recorded thereon a computer program for executing the yield prediction simulation method is disclosed.

[0019] According to the present invention, there is a technical effect of being able to predict the yield of a second cycle period based on process operation data of a first cycle period and simulate the yield prediction by reflecting the tag fluctuation amount.

[0020] According to the present invention, the first cycle period is divided into a plurality of segments based on the catalyst life and the resulting yield change, and key factor analysis, yield prediction, and tag variation analysis are performed for each segment, and a yield prediction simulation for the second cycle period is performed accordingly, thereby having a technical effect of increasing prediction accuracy over the entire cycle period.

[0021] In addition, according to the present invention, by selecting key factors according to process operating conditions and applying them to a prediction model, the accuracy of prediction of changes in catalyst activity can be increased, and by enabling product sales planning and catalyst replacement timing to be determined through accurate catalyst activity prediction, the time and cost required for product production and catalyst replacement can be efficiently managed, and by identifying changes in catalyst activity according to process operating conditions and optimizing operating conditions, the catalyst activity can be improved, thereby increasing product production.

[0022] Figure 1 is a block diagram of a yield prediction simulation system according to one embodiment of the present invention.

[0023] Figure 2 is a flowchart showing a yield prediction simulation method according to one embodiment;

[0024] Figure 3 is a flowchart illustrating a data preprocessing method according to one embodiment;

[0025] Figure 4 is a flowchart illustrating a segment analysis method according to one embodiment;

[0026] Figures 5 to 10 are drawings explaining a segment analysis process according to one embodiment.

[0027] Figures 11 and 12 are drawings illustrating a data realization method according to one embodiment;

[0028] Figure 13 is a drawing explaining a life factor according to one embodiment;

[0029] Figure 14 is a flowchart illustrating a method for calculating a life factor according to one embodiment.

[0030] Figure 15 is a drawing showing a yield prediction result according to one embodiment;

[0031] FIG. 16 is a drawing illustrating a user interface (UI) for tag variation analysis according to one embodiment;

[0032] Figure 17 is a diagram showing the results of a yield prediction simulation according to one embodiment.

[0033] Figure 18 is a drawing explaining the results of a yield prediction simulation reflecting a life factor according to one embodiment.

[0034] According to one embodiment of the present invention, a yield prediction simulation method for predicting a yield of a second cycle based on yield-related data of a first cycle in a chemical process is disclosed, the method comprising: a step of predicting the yield of the second cycle using a machine-learned yield prediction model using data of the first cycle as learning data; a tag variation analysis step of calculating a variation in at least some of the remaining input variables due to a variation in one or more of the input variables input to the yield prediction model as an operating condition of the chemical process; and a step of calculating a simulated yield of the second cycle based on the predicted yield calculated in the yield prediction step and the tag variation analysis result calculated in the tag variation analysis step.

[0035] The above-described purposes, other purposes, features, and advantages of the present invention will be readily understood through the following preferred embodiments, illustrated in the accompanying drawings. However, the present invention is not limited to the embodiments described herein and may be embodied in other forms. Rather, the embodiments introduced herein are provided to ensure that the disclosure is thorough and complete, and to ensure that the spirit of the present invention is fully conveyed to those skilled in the art.

[0036] When terms such as "first," "second," etc. are used herein to describe components, these components are not intended to be limited by these terms. These terms are used solely to distinguish one component from another. The embodiments described and illustrated herein also include complementary embodiments.

[0037] In this specification, the singular also includes the plural unless specifically stated otherwise. The expressions "comprises," "consists of," and "consist of" used in the specification do not exclude the presence or addition of one or more other components in addition to the components mentioned.

[0038] In this specification, the term 'software' means a technology that moves hardware in a computer, the term 'hardware' means a type of device or apparatus that constitutes a computer (CPU, memory, input device, output device, peripheral device, etc.), the term 'step' means a series of processes or operations connected in time series to achieve a certain goal, the term 'computer program', 'program', or 'algorithm' means a set of instructions suitable for processing by a computer, and the term 'program recording medium' means a computer-readable recording medium that records a program used to install, execute, or distribute a program.

[0039] The terms ‘part’, ‘module’, ‘unit’, ‘block’, ‘board’, etc. used in this specification to refer to components of the invention may mean a physical, functional, or logical unit that processes at least one function or operation, and which may be implemented by one or more hardware, software, or firmware, or a combination of hardware, software, and / or firmware.

[0040] In this specification, the terms "processing unit," "computer," "computing device," "server device," and "server" may be implemented as a system having an operating system such as Windows, Mac, or Linux, a computer processor, memory, application programs, and storage devices (e.g., HDD, SSD). The computer may be, for example, a desktop computer, a laptop, or a mobile terminal, but these are exemplary and not limited thereto. The mobile terminal may be one of a smartphone, a tablet PC, or a mobile wireless communication device such as a PDA.

[0041] The present invention will now be described in detail with reference to the drawings. In describing specific embodiments below, various specific details have been included to further illustrate the invention and facilitate understanding. However, readers with sufficient knowledge of the field will recognize that the present invention can be used without these specific details. Furthermore, it should be noted in advance that, in describing the invention, parts that are known or commonly used but not significantly related to the invention will not be described to avoid confusion in the description of the invention.

[0042] Figure 1 is a block diagram briefly illustrating a yield prediction simulation system according to one embodiment of the present invention. In the following description of the present specification, the yield prediction simulation system according to the present invention (hereinafter also simply referred to as the “yield prediction system” or “simulation system”) is explained assuming that it is applied to an olefin production process. For example, the simulation system of the present invention can be applied to a PDH (Propane DeHydrogenation) process that produces propylene using propane as a raw material, and through this process, hydrogen can be extracted from propane to produce propylene, a type of olefin.

[0043] In one embodiment of the present invention, a yield prediction simulation system can predict the yield of a second cycle period based on process operation data collected during a first cycle period. The first and second cycles may be of the same length, but may also be of different lengths. In the embodiments of this specification, a cycle period of four years is assumed as an example. Preferably, one cycle may be related to the lifespan of a catalyst used in a chemical process. For example, if the catalyst lifespan is four years, one cycle may be set to four years.

[0044] Referring to FIG. 1, a yield prediction simulation system (100) according to one embodiment may include a data preprocessing unit (110), a segment analysis unit (120), a data realization processing unit (130), a life factor analysis unit (140), a key factor analysis unit (150), a yield prediction unit (160), a tag variation analysis unit (170), and a yield prediction simulator (180), and each of these components (110 to 180) may be implemented as software that is executable and programmed on a computer device, and may also be implemented in combination with firmware and hardware as needed.

[0045] The data preprocessing unit (110) is a functional unit that collects and extracts data from the data storage unit (200) and preprocesses the data. The operation of the data preprocessing unit (110) will be described later with reference to FIG. 3.

[0046] The segment analysis unit (120) can divide a cycle into multiple segments based on preprocessed data. For example, for a cycle (e.g., 4 years), a cycle is divided into multiple periods based on the amount of change in certain factors, such as process temperature and yield, according to the lifespan (aging) of the catalyst used in the process. An exemplary operation of the segment analysis unit (120) will be described below with reference to FIGS. 4 to 10.

[0047] The data realization processing unit (130) is a functional unit that generates data used for yield prediction by realizing it into a form suitable for input into a yield prediction model. The yield prediction simulation system according to the present invention utilizes data from a past cycle (the first cycle) to predict a future cycle (the second cycle), and at this time, data for the second cycle can be generated based on the data from the first cycle and input into the yield prediction model. An exemplary operation of the data realization processing unit (130) will be described below with reference to FIGS. 11 and 12.

[0048] The life factor analysis unit (140) is a functional unit that reflects the life (aging) of the catalyst used in the process to more accurately predict the process yield. Generally, the life of a catalyst varies depending on the type, and the aging trend within the life period can also vary. In particular, if the catalyst ages rapidly toward the latter stage, it may be difficult to accurately reflect this in the yield prediction model. Therefore, in one embodiment of the present invention, the catalyst life factor (aging factor) is additionally considered. For example, the catalyst life factor over time can be calculated and this value can be reflected as a weight in the process data input to the yield prediction model, thereby improving the yield prediction performance. An exemplary operation of the life factor analysis unit (150) will be described below with reference to FIGS. 13 to 15 .

[0049] The key factor analysis unit (150) extracts process key factors using data preprocessed by the data preprocessing unit (110). For example, the key factor analysis unit (150) may be implemented using a known machine learning algorithm, such as a machine learning algorithm that utilizes a feature selection technique.

[0050] The yield prediction unit (160) is a functional unit that predicts the yield of a process using preprocessed data and extracted key factors. In one embodiment, the yield prediction unit (160) is implemented as a machine learning-based learning model, and can train a yield prediction model using preprocessed data and key factors, and predict the yield using the trained yield prediction model. For example, if the first cycle data and the first half of the second cycle data are preprocessed, the yield prediction model can be trained using the preprocessed data and key factors, and then the yield prediction results for the remaining period of the second cycle can be output.

[0051] The tag variation analysis unit (170) is a functional unit that calculates the variation of tags (input variables input into the yield prediction model). In the present invention, 'tags' are input variables input into the yield prediction model and mean various operating conditions such as temperature, pressure, and flow rate in a chemical process, for example. There are tags (hereinafter referred to as 'control tags') that can be controlled and adjusted by a user (e.g., a process operator or worker of a factory, etc.). When the worker changes the tag value of this control tag, the tag values ​​of the remaining tags also change. The tag variation analysis unit (170) can calculate the variation of at least some of the remaining tags due to a change in one or more of these control tags, and can be implemented, for example, with a machine learning algorithm trained with process operation data stored in the data storage unit (200).

[0052] The yield prediction simulator (180) is a functional unit that simulates changes in yield based on the yield predicted by the yield prediction unit (160) and the tag variation analysis result calculated by the tag variation analysis unit (170). While the yield prediction unit (160) statically predicts the yield of the second cycle period based on the data stored so far (data of only the first cycle, or data of the first cycle and the first half of the second cycle), the yield prediction simulator (180) can simulate how the future yield changes when the process conditions (temperature, pressure, flow rate, etc.) are changed. To this end, the yield prediction simulator (180) receives the predicted yield calculated by the yield prediction unit (160) and the analysis result calculated by the tag variation analysis unit (170) together as input to simulate the yield change.

[0053] An exemplary operation of a yield prediction simulation system will be described with reference to FIG. 2. FIG. 2 is a flowchart showing a yield prediction simulation method according to an embodiment. Referring to FIG. 2, the yield prediction simulation method according to an embodiment includes a step (S10) of preprocessing data for yield prediction including data of at least a first cycle, a segment analysis step (S20) of dividing one cycle into a plurality of segments based on the preprocessed data, a data realization step (S30) of generating and realizing data of a second cycle based on the first cycle data and the first half data of the second cycle, and a life factor analysis step (S40) of calculating a life factor of a catalyst used in a process.

[0054] In addition, the yield prediction simulation method includes a step (S50) of analyzing key factors for each segment based on the segment analysis results, and a step (S60) of modeling a yield prediction model for each segment to predict the yield of the second cycle.

[0055] In addition, the yield prediction simulation method may further include a tag change amount analysis step (S70) for calculating the change amount of the remaining tags when the user changes the tag value of the control tag, and a yield prediction simulation step (S80) for simulating the change amount of the yield by reflecting the changed tag values ​​in the predicted yield predicted in the yield prediction step (S60).

[0056] Hereinafter, each step of FIG. 2 will be described in more detail. FIG. 3 illustrates an exemplary method of a data preprocessing step (S10) according to one embodiment. Referring to FIG. 3, the data preprocessing step (S10) may include a step (S110) of preprocessing data for yield prediction in units of minutes, a step (S120) of selecting a tag to be analyzed, a step (S130) of extracting hourly and daily data for the data of the selected tag among the data preprocessed in units of minutes, and a step (S140) of performing outlier processing and missing value interpolation for the daily data.

[0057] In order to preprocess data in step (S110), data required for yield prediction is collected and extracted from the data storage unit (200). At this time, the data storage unit (200) may be implemented as a database, for example, but the data format is not particularly limited. In one embodiment, the data for yield prediction extracted from the data storage unit (200) may include (i) process operation data of an olefin production plant, (ii) laboratory data including LIMS data, (iii) plant event data including data regarding times when the plant was not operating normally, and (iv) past yield, conversion rate, and selectivity data regarding olefin production.

[0058] Process operation data from an olefin production plant (PDH plant) may be collected from sensors installed in various plant facilities (e.g., reactors, pipelines, etc.). Each sensor may measure variables that can monitor process operation conditions, such as temperature, pressure, flow rate, and composition, and data may be collected from each sensor on a minute-by-minute basis.

[0059] Process operation data can be classified by section, unit, and tag and stored in the data storage unit (200). At this time, a unit is a mid-size set of tags within the factory, a section is a large-size set of units, and multiple sections are gathered to form the entire PDH factory. Meanwhile, a tag can function as an identifier that identifies each sensor installed in the factory. That is, a unique tag is assigned to each sensor, and for example, if more than 9,000 sensors are installed in a PDH factory, there may be the corresponding number of tags. In the following specification, unless there is a particular concern of confusion, data output from a sensor corresponding to each tag is also referred to as a 'tag' or 'tag data'.

[0060] Laboratory data may include LIMS (Laboratory Information Management System) data. In one embodiment, laboratory data, as well as actual observation data (tag data), may be utilized for accurate yield prediction. Laboratory data may also be used to handle and interpolate outliers or missing values ​​in tag data. In an alternative embodiment, laboratory data may be omitted.

[0061] Factory event data may include, for example, data on times when operations were not performed normally (shutdown history), large integers / small integers, etc., and may be used when analyzing and processing outliers or missing values ​​in tag data. Past yield values ​​include past yield data on olefin production. In addition to past yields, conversion and selectivity values ​​may also be included. In the following specification, yield, conversion, and selectivity are collectively referred to as "yield" unless there is a risk of confusion, and from a machine learning perspective, yield (i.e., yield, conversion, and selectivity) are also referred to as "targets."

[0062] The data stored in the data storage unit (200) is continuously accumulated in predetermined cycle units and may be data from more than one cycle period from the present time. In this case, the cycle unit may be seconds or minutes. For example, data may be collected in 30-second units and then converted and stored in minutes and / or hours for analysis. However, these cycle units are exemplary and are not limited to a specific cycle. In addition, one cycle may be set in relation to the catalyst life, and in one embodiment, it may be set to 4 years, but it will be understood that this is exemplary.

[0063] The yield prediction data extracted from the data storage unit (200) is preprocessed as minute-unit data in step (S110). For example, if second-unit data is received from the data storage unit (200), it is converted to minute-unit data, and if outliers or missing values ​​occur, outlier processing and missing value interpolation are performed.

[0064] Next, in step (S120), the analysis target tags are selected. For example, all tag data collected from all sensors installed in the factory can be used to perform the key factor analysis and yield prediction described below. However, preferably, some tag data can be selected from the entire tag data, and the selected tag data can be used to perform the subsequent (e.g., after step S130) hourly / daily data extraction, key factor analysis, and yield prediction. In this case, the analysis target tags are selected in step (S120). For example, tags recognized as useful for analysis can be selected based on past research and the knowledge and experience of field engineers.

[0065] In one embodiment, the step of selecting the analysis target tags (S120) may be performed in advance before the step of preprocessing the minute-by-minute data (S110), and in this case, the minute-by-minute data preprocessing (S110) may be performed only for the tags selected as the analysis target.

[0066] When the analysis target tag is selected in step (S120), time-based data is extracted in step (S130) and then processed again to extract daily data. At this time, data integration can be performed by utilizing process data (tag data) and LIMS data. In an alternative embodiment, the step (S120) of selecting the analysis target tag can be performed after extracting the time-based data. In this case, time-based data can be extracted for all process data, and then daily data can be extracted only for the analysis target tag.

[0067] After extracting daily data in step (S130), data preprocessing is performed in step (S140). For example, data preprocessing includes outlier processing and missing value interpolation. In the case of outlier processing, outliers are selected and excluded or corrected, and only the refined values ​​are used as valid input values. In addition, missing value interpolation is performed for sections that have been selected and removed as outliers or sections where process data is missing due to factory shutdown. Missing value interpolation can be performed by generating new data through, for example, linear regression and distribution-based random number generation.

[0068] The yield prediction data preprocessed through the above steps can be organized and converted into a data format to be used in a machine learning model and then stored in the data storage unit (200) or any other arbitrary storage unit.

[0069] FIGS. 4 to 10 are drawings explaining an exemplary method of a segment analysis step (S20 of FIG. 2) according to one embodiment. FIG. 4 is a flowchart explaining an exemplary method of a segment analysis step, and FIGS. 5 to 10 are drawings explaining a segment analysis process according to one embodiment.

[0070] In the segment analysis step (S20), a cycle is divided into multiple segments based on preprocessed data. In one embodiment, by dividing the segments by considering the inflection point where the yield trend changes rapidly and the degree of yield fluctuation, the segments can be divided into sections showing similar yield increase / decrease trends during the process operation period of a cycle. By utilizing each segment for various modeling such as subsequent key factor extraction, yield prediction, and yield prediction simulation, the accuracy of yield prediction can be improved.

[0071] However, in an alternative embodiment, the segment analysis step (S20) may be omitted, and in this case, it is of course possible to use each model of key factor extraction, yield prediction, and yield prediction simulation for one cycle, without dividing the cycle into multiple segments.

[0072] Referring to FIG. 4, in one embodiment, the segment analysis step (S20) may include a step of selecting key factors required for segment analysis (S210), a step of selecting an inflection point of the key factors (S220), and a step of determining segments by integrating or separating segments through volatility analysis of the key factors (S230). In addition, in one embodiment, after step (S230), a step of confirming or modifying segments based on catalyst design (S240) may be further included.

[0073] Briefly explaining each step, first, in step (S210), key factors required for segment analysis are selected. For example, key factors can include at least one of the target values ​​(yield, conversion rate, selectivity), as well as key tags that influence these target values. In this regard, Figure 5 exemplifies the key factors selected in step (S210).

[0074] Next, in the step of selecting an inflection point (S220), the inflection point of the main factor is selected. As a specific method of selecting an inflection point, (i) inflection points for one or more factors required for segment selection within a cycle are calculated, (ii) the calculated inflection points are clustered, and (iii) among the clustered inflection points, an inflection point that serves as a segment boundary is selected.

[0075] More specifically, in the step (i) above, the inflection point can be calculated using a known method such as the Plateau Detection method. As an example, Fig. 6 shows the result of detecting the inflection point using the modified Plateau Detection method. In Fig. 6, the graph is a graph of air temperature during catalyst regeneration (Regen Air Temperature), and the X-axis represents time (or a variable corresponding to time such as 'cumulative production volume') within one cycle, and the Y-axis represents temperature.

[0076] Thereafter, in the step (ii) above, inflection points are clustered. For example, Fig. 7 shows clustering of inflection points using the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) method, which clusters data using the degree of data density. DBSCAN is one of the known clustering methods, and the present invention is not limited to this method.

[0077] Next, in the above step (iii), the segments are initially determined by selecting the inflection points that serve as segment boundaries among the clustered inflection points. For example, Fig. 8 shows a state in which four inflection points (red points) have been selected by the above step (iii).

[0078] Figures 6 to 8 illustrate exemplary methods for selecting inflection points for Regen Air temperature among major factors. Multiple inflection points can be selected for other major factors such as yield, selectivity, and conversion rate in the same or similar manner. For example, Figure 9(a) shows the results of selecting four inflection points (red points) for yield, and Figure 9(b) shows the results of selecting four inflection points (red points) for selectivity.

[0079] In one embodiment, segments are determined by integrating or separating segments through volatility analysis using multiple inflection points selected for each major factor as described above (step S230). In step (S230), for example, for at least some major factors, the segment-specific average and deviation are calculated for each segment distinguished by the inflection points selected in step (S220), and the segments are integrated or separated based on the calculated average and deviation to determine the final segment.

[0080] In one embodiment, the segmentation result determined as above can be used to proceed to the next step (e.g., the key factor analysis step (S50) of FIG. 2, etc.), and alternatively, after step (S230), a step (S240) of confirming and modifying segments based on the catalyst productivity design can be further included. In step (S240), for example, the segment point is confirmed through a comparison of similarity with the catalyst design (catalyst characteristics, lifespan, etc.) taking into account the cumulative product production volume, and if necessary, the segmentation is finalized by modifying it.

[0081] Figure 10 illustrates the segmentation results finally determined by this segment analysis step (S20). In Figure 10, when the catalyst life is set to one cycle (e.g., 4 years), one cycle is divided into four segments (SG1 to SG4). In Figure 10, the red graph represents the air temperature during catalyst regeneration (Regen Air Temperature), the gray graph represents the yield, and the light green graph represents the selectivity.

[0082] In Fig. 10, the first segment (SG1) is a period in which the catalyst is injected, the regen air temperature gradually increases, and the process moves toward stabilization, and the second segment (SG2) is a stabilization period in which the regen air temperature is maintained stably and the yield and selectivity are stable. In the third segment (SG3), the yield is seen to gradually decrease as the catalyst ages. In other words, even if the regen air temperature is increased, the yield is not maintained or increased due to catalyst aging, but gradually decreases. The fourth segment (SG4) is a period in which the yield decreases more rapidly, and even if the regen air temperature is increased further, the yield and selectivity no longer increase but rapidly decrease.

[0083] In this way, in the present invention, by considering the increase / decrease trend of the yield according to the catalyst life within one cycle period, it is possible to divide it into segments showing similar trends, and for each divided segment, in subsequent steps (e.g., the key factor analysis step (S50), the yield prediction step (S60), the yield prediction simulation step (S80) of FIG. 2, etc.), modeling is performed for each segment, and the analysis / prediction results are derived continuously for one entire cycle, thereby increasing the accuracy of yield prediction.

[0084] Hereinafter, an exemplary method of the data actualization step (S30 of FIG. 2) will be described with reference to FIGS. 11 and 12. The data actualization step (S30) generates and actualizes data for the remaining period of the second cycle based on the first cycle data and the first half of the second cycle data. The yield prediction step (S60) according to the present invention predicts a future cycle (the second cycle) by utilizing data of a past cycle (the first cycle). More specifically, the yield of the second cycle is predicted by generating data for the second cycle based on the data of the first cycle and then inputting the second cycle data into a yield prediction model. At this time, if there is data for a portion of the second cycle (hereinafter also referred to as the “first half of the second cycle”), the data actualization step (S30) generates data for the remaining period of the second cycle by utilizing the data of the first half of the second cycle.

[0085] For example, assuming that one cycle period is 4 years, the first cycle is from January 2017 to December 2020, and the second cycle is from January 2021 to December 2024. Assuming that the current point in time is August 2022, the data storage unit (200) stores process operation data of the first cycle and data for the first half of the second cycle (i.e., from January 2021 to July 2022).

[0086] In this case, the data realization processing unit (130) generates data for the remaining period of the second cycle (i.e., from August 2022 to December 2024) based on the process operation data of the first cycle and the data for the first half of the second cycle. At this time, the data realization processing unit (130) may generate data for the remaining period of the second cycle by considering characteristics such as trends or averages of the data for the first cycle and characteristics of the data for the first half of the second cycle.

[0087] However, the process operation data (e.g., each tag data) of the first and second cycles tend to have different trends or values ​​between the two cycles due to differences in catalyst input amount, initial operation conditions, etc., and therefore, there is a problem in that it is difficult to generate data for the second cycle by directly applying the operation condition data of the first cycle. Therefore, in the data realization step (S30) of the present invention, the data for the second cycle is generated by correcting the data for the first cycle to suit the trend of the second cycle.

[0088] In one embodiment, the method of realizing each tag data may include at least a mean difference reflection method and a random number generation method.

[0089] The average difference reflection method can be applied when there is an average difference between the first cycle and the second cycle. In one embodiment, when the data within a predetermined period of the first cycle and the second cycle have similar fluctuation trends but different average values, the average point of the first cycle data is moved to generate the data of the second cycle. For example, Fig. 11 shows exemplary tag data to which the average difference reflection method can be applied. In Fig. 11, the X-axis corresponds to the time of one cycle, and the Y-axis represents the data value of the corresponding tag. In addition, the black graph is the data of the first cycle (CY1) of the corresponding tag, and the red graph is the data of the first half of the second cycle (CY21). It will be understood that the current point in time is the end point of the data of the first half of the second cycle (CY2).

[0090] When comparing the first cycle data (CY1) with the first half of the second cycle data (CY21), the trends of the two data are similar, but the average value of the second cycle data appears to be larger. Therefore, in this case, the average difference reflection method can be applied to generate data for the remaining period of the second cycle by increasing the data of the corresponding period of the first cycle by the average difference. In this case, in one embodiment, the data of the corresponding period of the first cycle can be used as is with the average increased, or alternatively, the second data can be generated by modifying the data by a method such as random number generation for at least some sections.

[0091] Random number generation can be applied when data for a given period is incomplete or contains outliers. In one embodiment, if incomplete data exists in the first cycle, a random number is generated to generate data for the second cycle. For example, Figure 12 illustrates exemplary tag data to which random number generation can be applied.

[0092] In Figure 12, the X-axis corresponds to the time of one cycle, and the Y-axis represents the data value of the corresponding tag. The black graph represents the data of the first cycle (CY1) of the corresponding tag, and the orange graph represents the data of the first half of the second cycle (CY21). The end point of the data of the first half of the second cycle (CY2) represents the current point in time.

[0093] Referring to Fig. 12, the first cycle data (CY1) was generated after a certain point in time. In other words, the tag data may mean that the sensor was not installed or was not operating before the certain point in time. However, for the second cycle, the first half data (CY21) exists, and the remaining section data (CY22) is generated by generating a random number based on the data of the first cycle to generate the second half data (CY22). In this case, for example, the second half data (CY22) can be generated by calculating the average and variance of a certain section in the past from the current point in time (for example, 30 days in the past (D30) from the current point in time) and generating a random number for data at individual times while maintaining this average and variance.

[0094] Now, an exemplary method of the life factor analysis step (S40 of FIG. 2) will be described with reference to FIGS. 13 to 16. The life factor analysis step (S40) may be performed to reflect the aging of the catalyst used in the process in order to more accurately predict the simulation yield in the yield prediction simulation step (S80). For example, in a PDH process for producing propylene, as the cumulative production of propylene increases, the catalyst life decreases, causing the catalyst yield to drop sharply in the latter half of the process. In order to accurately reflect not only the yield change due to changes in process conditions but also the yield decrease due to the catalyst life in the simulation yield prediction, it may be desirable to perform simulation yield prediction by applying the catalyst life factor.

[0095] For example, referring to FIG. 13, the yellow graph represents the yield (Y1) of the first cycle over time in one cycle, and the red graph represents the life factor (AF1) of the first cycle. It can be seen that the life factor (AF1) increases and decreases in a similar trend to the yield (Y1). In addition, as explained with reference to FIG. 10, in the fourth segment, which is the last section of the cycle, the yield decreases no matter how much the temperature is increased. At this time, the yield decreases relatively linearly up to the third segment, but in the fourth segment, the yield decreases nonlinearly.

[0096] Meanwhile, the green graph represents the yield (Y2) up to the current point in the second cycle, and the blue graph represents the life factor (AF2) up to the local point in the second cycle. They are progressing in a similar trend to the yield (Y2) and life factor (AF1) of the first cycle, respectively. Therefore, it can be estimated that the yield (Y2) also decreases nonlinearly for the fourth segment period of the second cycle. In order to more accurately predict the simulation yield in this period, the simulation yield is calculated by reflecting the life factor (AF) for the second cycle.

[0097] At this time. In one embodiment, since the yield prediction model predicts the yield relatively accurately for the first and second segment sections during the entire period of the second cycle without applying the life factor (AF), the life factor (AF) may be applied to the third and fourth segment sections without applying the life factor to the first and second segments. In another embodiment, since the yield of the third segment decreases relatively linearly, the yield prediction model can predict it to a certain degree of accuracy, the life factor (AF) may be applied only to the fourth segment. As another example, the time point for applying the life factor may be determined based on the cumulative propylene production amount and / or the life factor value. That is, the life factor may be applied from the time point when the cumulative propylene production amount becomes equal to or greater than a preset predetermined production amount, or the life factor may be applied from the time point when the life factor value becomes equal to or less than a preset predetermined value.

[0098] Figure 14 is a flowchart illustrating a method for calculating a life factor according to one embodiment.

[0099] Referring to Figure 14, the catalyst aging factor is calculated in step S410. In one embodiment, the aging factor can be calculated based on the production volume relative to the amount of heat applied to the catalyst. For example, the catalyst aging factor is calculated by dividing the daily propylene production volume by the amount of heat applied to the catalyst, as shown in Equation 1 below.

[0100] AF (Aging Factor) = (Daily propylene production) / (Heat applied to catalyst) --- Formula 1

[0101] In the above formula 1, the “heat applied to the catalyst” can be calculated by multiplying the air temperature (Regen Air temperature) and the flow rate during catalyst regeneration, for example.

[0102] In step S420, the calculated life factor is divided into multiple time intervals. For example, the life factor may be divided into intervals based on the cumulative propylene production volume over a typical temperature change cycle of two weeks.

[0103] Next, weights to be applied to the simulation yield are calculated from the life factors in each of the above-mentioned segments. A specific method for calculating the weights is, for example, as shown in FIG. 14, to normalize the life factors of each segment (step S430) and calculate the weights using the normalized life factors (step S440). In this embodiment, in step S430, a representative value of the life factor is selected for each segment and normalized. The representative value may be, for example, the average value of each segment, but this is exemplary and the representative value may be selected for each segment using any other arbitrary method. After calculating the representative value for each segment, the life factors are normalized to have values ​​between 0 and 1, for example, using the Min-Max normalization method.

[0104] In the subsequent step (S440), the normalized lifetime factor is scaled to calculate the lifetime factor weight to be applied to the simulation yield. The lifetime factor weight is a value that is multiplied by the simulation yield output as a result of the yield prediction simulation (S80 in Fig. 2), and is a value in which the lifetime factor is scaled in consideration of the extent to which the simulation yield is further reduced. For example, if it is determined that it is appropriate to reduce it within 1% of the simulation yield, the lifetime factor weight will have a value between 0.99 and 1.00. That is, through step (S440), the normalized lifetime factor is converted into a value (life factor weight) scaled between 0.99 and 1.00. However, this weight range is exemplary, and it goes without saying that the weight range can be set differently depending on the specific implementation situation of the invention.

[0105] In one embodiment, the life factor weight (C) is calculated according to Equation 2 below.

[0106] Life Factor Weight (C) = Normalized Life Factor (A) * Reflection Ratio (B) + (1-Reflection Ratio (B)) --- Formula 2

[0107] In the above formula 2, the normalized life factor (A) is the life factor calculated through step (S430). The reflection ratio (B) is a value determined according to the range to which the life factor weight (C) is scaled. For example, if the final simulation yield value is reduced within 1% of the simulation yield as above, the reflection ratio (B) becomes 0.01. Therefore, the normalized life factor (A) and reflection ratio (B) calculated for each section can be input into the above formula 2 to calculate the scaled life factor, i.e., the life factor weight (C), for each section.

[0108] As described above, after calculating the life factor weight (C) through steps (S410 to S440), when performing a yield prediction simulation as indicated in step (S810), the weight (C) is applied to the life factor application section to obtain the final simulation yield. That is, when executing step (S80) of Fig. 2, the life factor is reflected in the yield from the point in time when the predetermined criteria for applying the life factor are met, and this will be described later with reference to Fig. 18.

[0109] Now, referring back to FIG. 2, we will briefly describe the key factor analysis step (S50). The key factor analysis step (S50) extracts process key factors that affect yield using data preprocessed in the data preprocessing unit (110). For example, key factors for the second cycle data generated in the data realization step (S30) can be extracted for each segment divided by the segment analysis step (S20). The key factor extraction method can be implemented using a machine learning algorithm that utilizes a feature selection technique, such as a known machine learning algorithm, such as the Boruta algorithm.

[0110] The key factors extracted in step (S50) can be utilized in the yield prediction model in the subsequent yield prediction step (S60). For example, in this step (S50), about 100 key factors can be selected from approximately 9,000 tags in the factory, and in the subsequent yield prediction step (S60), yield prediction can be performed based on the tag data values ​​of the 100 or so selected key factors. Furthermore, in one embodiment, these key factors can also be utilized in the tag variation analysis step (S70).

[0111] Referring to Figure 2, the yield prediction step (S60) uses preprocessed data and key factors to predict the process target (at least one of yield, conversion rate, and selectivity). For example, the yield of the second cycle can be predicted using the second cycle data generated in the data realization step (S30) for each segment divided by the segment analysis step (S20).

[0112] For example, a yield prediction model can be used by ensembling a bagging series algorithm such as Random Forest and a boosting series algorithm such as XGBoost and LightGBM (LGBM) to create a yield prediction model with high predictive power, and the model can be trained using the first cycle data as training data.

[0113] Bagging algorithms, which utilize multiple iterations of random sampling in parallel to aggregate data, have the advantage of increasing the training data. Therefore, even when training data is insufficient, they provide sufficient learning effects, effectively preventing underfitting and overfitting. Boosting algorithms also perform multiple random samplings, but they perform them sequentially rather than in parallel. They adjust the weights of the next training data based on the previous training results, thereby progressing learning. This means that incorrect answers are given a higher weight, resulting in higher accuracy. Among these models, it may be desirable to apply the Random Forest algorithm, a widely used bagging algorithm that prevents overfitting and underfitting, or the LGBM model, a boosting algorithm that improves accuracy by learning errors. However, these yield prediction models are exemplary, and any suitable known machine learning method can be used depending on the specific embodiment of the invention. Figure 15 is an exemplary screen configuration showing the results of yield prediction using the ensemble model described above, and in Figure 15, the predicted yield (Y) during the second cycle period is shown as an orange graph.

[0114] Referring to Fig. 16, the tag variation analysis step (S70 of Fig. 2) is described. In the tag variation analysis step (S70), the tag variation is analyzed using a tag variation analysis model that calculates the variation for at least some of the remaining tags due to a change in one or more control tags among the yield prediction data.

[0115] In one embodiment, at least one control tag that can be manipulated is selected from among tags used for yield prediction. Here, the 'control tag' means a tag that can be manipulated by a user among process conditions of a chemical process, and may include, for example, at least one tag among air temperature during catalyst regeneration (Regen Air temperature), raw material heating temperature (Charge Heater temperature), air flow rate during catalyst regeneration (Regen Air flow rate), and reactor raw material injection flow rate (Reactor Feed flow rate). Therefore, it will be understood that when, for example, 100 factors (tags) are used for yield prediction, the four tags become control tags, and at least some of the remaining 96 tags become tags that change due to changes in the four control tags.

[0116] The tag variation analysis unit (170) can calculate the variation of at least some of the remaining tags due to the variation of one or more control tags, and can perform tag variation analysis (S70) on the tags selected by the second cycle data generated in the data realization step (S30) and the key factor analysis (S50) for each segment divided by the segment analysis step (S20), for example.

[0117] In this regard, Fig. 16 shows an exemplary user interface (UI) for tag variation analysis. In the embodiment of Fig. 16, four tags are used as control tags, namely, air temperature during catalyst regeneration (Regen Air Temperature), raw material heating temperature (Charge Heater Temperature), air flow rate during catalyst regeneration (Regen Air Flow Rate), and reactor feed flow rate (Reactor Feed Flow Rate). For example, an arrow button (10) may be displayed on the right side of the Regen Air Temperature tag so that the operator can increase or decrease the tag value thereof, and it will be understood that the tag values ​​of the remaining three control tags can also be adjusted with the arrow buttons, respectively. In addition, when the operator adjusts at least one of the four control tags to an arbitrary value and then presses the 'Start Analysis' button (20) below (for example, clicks with the mouse), the tag variation analysis model can calculate and output the variation of the remaining tags according to the change in the control tag.

[0118] And the tag variation analysis result performed in the tag variation analysis step (S70) as above can be used as an input variable in the subsequent yield prediction simulation step (S80). The yield prediction simulation step (S80) is a step that simulates the change in yield based on the yield predicted by the yield prediction unit (160) and the tag variation analysis result produced by the tag variation analysis unit (170) and outputs the simulated yield. In other words, while the yield prediction step (S60) statically predicts the future yield based on the data of the second cycle generated by the data actualization (S30), the yield prediction simulation step (S80) dynamically predicts how the future yield changes when the user changes the control tag to an arbitrary value through the tag variation analysis (S70) and the remaining tag values ​​are also changed accordingly. Therefore, in the yield prediction simulation step (S80), the yield prediction simulator (180) inputs the predicted yield produced by the yield prediction unit (160) and the analysis results produced by the tag variation analysis unit (170) together into the yield prediction simulation model to produce a simulation yield.

[0119] The yield prediction simulation model can be implemented using a machine learning algorithm, and for example, a prediction model with high predictive power can be used by ensembling bagging algorithms such as Random Forest and boosting algorithms such as XGBoost and LGBM.

[0120] Fig. 17 shows exemplary yield prediction simulation results according to one embodiment. In Fig. 17, the orange graph is the predicted yield (Y) in the yield prediction step (S60), which is the same as the yield graph of Fig. 15. In Fig. 17, the pink graph is the simulated yield (“simulated yield”) (“Ys”) in the yield prediction simulation step (S80). That is, it shows the simulated yield in a state where an operator changes at least one control tag value on the screen of Fig. 16 and the remaining tag values ​​are also changed accordingly. In this way, according to the present invention, not only static yield prediction is performed using the second cycle data, but also a dynamic simulation is performed to show how the yield changes due to the change in the control tag when the main control tag is changed, thereby performing and predicting yield analysis due to the change in the main tag more accurately and precisely.

[0121] Additionally, in one embodiment, when calculating the simulation yield, the life factor may be applied as a weight for a predetermined period of the second cycle (e.g., the third and fourth segment periods of the second cycle, or the fourth segment period, etc.).

[0122] As shown in step (S810) in Fig. 14, after calculating the life factor weight (C) through steps (S410 to S440), when performing a yield prediction simulation, the weight (C) is applied to the life factor application section to obtain the final simulation yield.

[0123] For example, in this step (S810), the final simulation yield (Ysf) is calculated as shown in Equation 3 below.

[0124] Final simulation yield (Ysf) = simulation yield (Ys) * weight (C) --- Equation 3

[0125] In the above formula, the simulation yield (Ys) means the simulation yield calculated based on the predicted yield by the yield prediction step (S60) and the tag variation analysis result by the tag variation analysis (S70), and the weight is the life factor weight (C) calculated according to formula 2.

[0126] Figure 18 schematically illustrates the simulation yield that reflects the life factor. In Figure 18, the horizontal axis represents the cumulative propylene production volume and the vertical axis represents the simulated yield. The black graph in the figure represents the non-simulated yield, i.e., the predicted yield output from the yield prediction step (S60). The red solid line graph represents the simulated yield output from the yield prediction simulation step (S80) when the temperature is increased among the control tags, and the blue solid line graph represents the simulated yield output when the temperature is decreased, each graph not reflecting the life factor.

[0127] In one embodiment, the life factor can be applied from a point in time when both the condition that the cumulative propylene production amount is greater than or equal to a predetermined production amount (P1) and the condition that the life factor value is less than or equal to a predetermined value (AF1) are satisfied. For example, in FIG. 18, in the case of the simulation yield (red graph) in the temperature increase simulation, from the point in time when the cumulative production amount becomes P1, the simulation yield (Ys) is multiplied by a weight (C) as in Equation 3 to output the final simulation yield (Ysf) (red dotted line), and in the temperature decrease simulation, from the point in time P2 when the life factor becomes AF1, the weight (C) is applied to output the final simulation yield (Ysf) (blue dotted line). By considering the catalyst life in this way, simulation predictions closer to the actual yield become possible.

[0128] As described above, those skilled in the art will appreciate that various modifications and variations are possible based on the description contained herein. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the claims set forth below but also by equivalents thereof.

[0129] (Explanation of symbols)

[0130] 100: Yield Prediction Simulation System 110: Data Preprocessing Unit

[0131] 120: Segment Analysis Department 130: Data Realization Processing Department

[0132] 140: Life Factor Analysis Department 150: Key Factor Analysis Department

[0133] 160: Yield Prediction Department 170: Tag Variance Analysis Department

[0134] 180: Yield Prediction Simulator 200: Data Storage

[0135] The present invention relates to yield prediction of a chemical process, and more specifically, to a yield prediction simulation method capable of predicting the yield of a second cycle period based on process operation data of a first cycle period and simulating the yield prediction by reflecting tag fluctuations.

Claims

1. A yield prediction simulation method for predicting the yield of a second cycle based on yield-related data of a first cycle in a chemical process. A step (S60) of predicting the yield of the second cycle using a machine-learned yield prediction model using the data of the first cycle as learning data; A tag variation analysis step (S70) for calculating the amount of change in at least some of the remaining input variables due to a change in one or more of the input variables input to the yield prediction model as an operating condition of the chemical process; and A yield prediction simulation method characterized by comprising a step (S80) of calculating a simulation yield of a second cycle based on the predicted yield calculated in the yield prediction step and the tag variation analysis result calculated in the tag variation analysis step.

2. In paragraph 1, A yield prediction simulation method characterized in that the first and second cycles are determined according to the life of a catalyst used in a chemical process.

3. In paragraph 1, For at least a portion of the second cycle, further comprising a life factor analysis step (S40) for calculating a life factor reflecting the life of the catalyst used in the chemical process, A yield prediction simulation method, characterized in that the above simulation yield calculation step (S80) calculates the simulation yield based on the predicted yield, the tag variation analysis result, and the life factor for at least a portion of the second cycle.

4. In paragraph 3, A yield prediction simulation method characterized in that the above life factor is calculated by a formula including (daily propylene production amount) / (heat applied to catalyst).

5. In paragraph 4, The above simulation yield calculation step (S80) is a yield prediction simulation method characterized in that, for at least a portion of the second cycle, the simulation yield calculated based on the predicted yield and the tag fluctuation analysis result is multiplied by a weight calculated from the life factor to calculate the final simulation yield.

6. In the third paragraph, the life factor analysis step (S40) is Step (S410) of calculating a life factor based on the production amount compared to the amount of heat applied to the catalyst; A step (S420) of dividing the calculated life factor into predetermined time intervals; and A yield prediction simulation method characterized by comprising a step of calculating a weight from a life factor in each of the above-mentioned segments.

7. In the 6th paragraph, the step of calculating the weight is: A step (S430) of normalizing the average life factor within each section above; and A yield prediction simulation method characterized by including a step (S440) of scaling the normalized life factor.

8. A computer-readable recording medium having recorded thereon a computer program for executing a yield prediction simulation method according to any one of claims 1 to 7.

Images