A machine learning model that incorporates the physical characteristics of adsorbents for post-combustion carbon capture.

Abstract

A power generation system is provided which includes a capture system for use in capturing carbon dioxide, a controller configured to operate the capture system, and a modeling system including a processor. The processor is configured to identify a model training dataset, where each instance of the model training dataset identifies an adsorbent used by the capture system, a carbon capture performance value for the adsorbent, and a plurality of primary feature values ​​associated with a plurality of primary features; to generate one or more secondary features based on one or more of the plurality of primary features; to generate a transfer function; and to use the transfer function to determine one or more promising adsorbents to be used by the capture system. The controller operates the capture system using the one or more promising adsorbents determined by the modeling system.

Description

[Technical Field]

[0001] This application relates in general to modeling post-combustion carbon capture systems, and more specifically to modeling the expected performance of novel adsorbents, specifically metal-organic frameworks, in post-combustion carbon capture systems, and to a system and method for operating a post-combustion carbon capture system using one or more promising adsorbents based on carbon capture performance values ​​determined by the model. [Background technology]

[0002] Some power plant systems may include post-combustion carbon capture ("PCC") systems configured to capture carbon dioxide (CO2) from the exhaust gases (flue gases) produced. PCC systems may be used to capture CO2 from exhaust gases produced by power plants, including, for example, coal combustion systems, gas turbines, and / or boilers. Some types of PCC systems use metal-organic frameworks ("MOFs") to facilitate carbon capture. Metal-organic frameworks typically consist of two main components: an inorganic metal component (often called a secondary construction unit, or "SBU") and an organic component (often called a "linker"). Various different MOFs have been developed and tested for their capture capacity (e.g., a measure of how effectively a particular MOF works in capturing CO2). However, there are potentially millions of combinations of metals, linkers, and other functional groups that can be used in MOFs, some of which may result in greater capture capacity and productivity than existing MOFs, and the actual creation and research of each possible combination of components is economically very expensive.

[0003] Systems and methods are needed to model MOF performance using known MOFs in order to predict how other proposed MOFs can function, to adjust the operation of carbon capture systems, and to use MOFs based on predicted carbon capture performance. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] International Publication No. 2021 / 067220 [Overview of the project]

[0005] In one embodiment, a power generation system is provided. The power generation system includes a capture system for use in capturing carbon dioxide, a controller configured to operate the capture system, and a modeling system including a processor. The processor is configured to identify a model training dataset, each instance of the model training dataset identifying an adsorbent used by the capture system, a carbon capture performance value for the adsorbent, and a plurality of primary feature values ​​associated with a plurality of primary features. The processor is also configured to generate one or more secondary features based on one or more of the plurality of primary features, each of which is a combination of at least two of the plurality of primary features, and to determine a plurality of correlation intensity values, including correlation intensity values ​​between each of the plurality of primary features and each of the one or more secondary features. The processor is also configured to identify a first subset of the model training dataset based on the plurality of correlation intensity values, determine the instance-by-instance statistical significance of the first subset of the model training dataset, and identify a second subset of the model training dataset based on the statistical significance, where the instance-by-instance statistical significance of the second subset of the model training dataset is below a predetermined threshold. The processor is further configured to generate a transfer function based on a second subset of the model training dataset and to use the transfer function to determine one or more potential adsorbents to be used by the capture system based on carbon capture performance values. The controller operates the capture system using one or more potential adsorbents determined by the modeling system.

[0006] In another embodiment, a method for selecting one or more promising adsorbents for use in operating a capture system to capture carbon dioxide. The method comprises identifying a model training dataset, where each instance of the model training dataset identifies an adsorbent to be used by the capture system, a carbon capture performance value of the adsorbent, and a plurality of primary feature values ​​associated with a plurality of primary features. The method also comprises generating one or more secondary features based on one or more of the plurality of primary features, where each of the one or more secondary features is a combination of at least two of the plurality of primary features; determining a plurality of correlation intensity values, where each of the plurality of primary features is a correlation intensity value; and identifying a first subset of the model training dataset based on the plurality of correlation intensity values. The method also comprises determining the instance-by-instance statistical significance of the first subset of the model training dataset; identifying a second subset of the model training dataset based on the statistical significance, where the instance-by-instance statistical significance of the second subset of the model training dataset falls below a predetermined threshold; and generating a transfer function based on the second subset of the model training dataset. The method further includes using a transfer function to determine one or more potential adsorbents to be used by the capture system based on carbon capture performance values, and the controller operates the capture system using one or more potential adsorbents determined using the transfer function.

[0007] The subject matter of this disclosure is described in more detail below with reference to exemplary embodiments shown in the accompanying drawings. [Brief explanation of the drawing]

[0008] [Figure 1] This figure shows an exemplary PCC modeling system that can be used to predict how a particular MOF (Mechanical Object Formation) can function in capturing carbon dioxide (CO2) from the exhaust gases of a coal-fired power plant. [Figure 2]A graph showing the decrease in CO2 adsorption productivity for an exemplary adsorbent (e.g., an exemplary MOF) when proceeding from a synthetic powder measured under equilibrium conditions to a final film coating on a substrate measured under dynamic conditions. [Figure 3A] A flowchart showing an exemplary method for analyzing the expected performance of a prospective adsorbent in a post-combustion carbon capture system. [Figure 3B] A flowchart showing an exemplary method for analyzing the expected performance of a prospective adsorbent in a post-combustion carbon capture system. [Figure 3C] A flowchart showing an exemplary method for analyzing the expected performance of a prospective adsorbent in a post-combustion carbon capture system. [Figure 4] A diagram showing an exemplary correlation matrix.

DETAILED DESCRIPTION OF THE INVENTION

[0009] The reference signs used in the drawings and their meanings are listed in summary form in a list of reference signs. In principle, the same reference sign is attached to the same component in the figure.

[0010] In the following specification and claims, several terms are referred to and are defined to have the following meanings.

[0011] As used herein, the singular forms "a", "an", and "the" include plural references unless the context clearly dictates otherwise. The terms "comprising", "including", and "having" are intended to be inclusive and mean that additional elements may exist other than the recited elements. The term "optional" or "optionally" means that the event or situation described thereafter may or may not occur, and that the description includes both the case where the event occurs and the case where it does not occur.

[0012] Unless otherwise indicated, the approximation terms used herein, such as “generally,” “substantially,” and “about,” indicate that the terms thus modified may apply only to an approximate degree as recognized by those skilled in the art, and not to an absolute or complete degree. Thus, values ​​modified by one or more terms such as “about,” “approximately,” and “substantially” are not limited to the exact values ​​specified. In at least some instances, the approximation terms may correspond to the precision of an instrument used to measure a value. Limitations of scope may be identified here and throughout this specification and the claims. Such scopes may be combined and / or substituted and include all sub-scopes contained herein unless the context or wording indicates otherwise.

[0013] Furthermore, unless otherwise indicated, terms such as “first,” “second,” etc., are used solely as labels in this specification and do not impose any order, position, or hierarchical requirements on the items referred to by these terms. Moreover, a reference to an item “second,” for example, does not require or exclude the presence of an item “first” or a lower-numbered item, or an item “third” or a higher-numbered item.

[0014] In one embodiment, a computer program is provided, and the program is implemented on a computer-readable medium. In an exemplary embodiment, the system runs on a single computer system without requiring a connection to a server computer. In a further embodiment, the system runs on a Windows® environment (Windows is a registered trademark of Microsoft Corporation, Redmond, Washington). In yet another embodiment, the system runs on a mainframe environment and a UNIX® server environment (UNIX is a registered trademark of X / Open Company Limited, located in Reading, Berkshire, United Kingdom). In yet another embodiment, the system runs on an iOS® environment (iOS is a registered trademark of Cisco Systems, Inc., located in San Jose, CA). In yet another embodiment, the system runs on a macOS® environment (macOS is a registered trademark of Apple Inc., located in Cupertino, CA). In yet another embodiment, the system runs on the Android® OS (Android is a registered trademark of Google, Inc., Mountain View, CA). In another embodiment, the system runs on the Linux® operating system (Linux is a registered trademark of Linus Torvalds of Boston, MA). The application is flexible and designed to run in a variety of different environments without compromising any major functionality. In some embodiments, the system includes multiple components distributed across multiple computing devices. One or more components may be in the form of computer executable instructions embedded in computer-readable media.

[0015] Any reference to “exemplary embodiments” or “one embodiment” in this disclosure used herein shall not be construed as excluding the existence of further embodiments that also incorporate the enumerated features.

[0016] As used herein, the terms “software” and “firmware” are interchangeable and include any computer program stored in memory for execution by a processor, including RAM memory, ROM memory, EPROM memory, EEPROM memory, and non-volatile RAM (NVRAM) memory. The memory types described above are merely examples and are therefore not limited to the types of memory that can be used to store computer programs.

[0017] The processors used herein may include any programmable system, including microcontrollers, reduced instruction set circuits (RISC), application-specific integrated circuits (ASICs), logic circuits, and any other circuits or processors capable of performing the functions described herein. The examples above are merely illustrative and therefore do not limit in any way the definition and / or meaning of the term “processor.”

[0018] The systems and processes are not limited to the specific embodiments described herein. In addition, each system and process component can be practiced independently of other components and processes described herein. Each component and process can also be used in combination with other assembly packages and processes.

[0019] As used herein, the term “non-temporary computer-readable medium” is intended to represent any tangible computer-based device implemented in any way or technique for short-term and long-term storage of information such as computer-readable instructions, data structures, program modules and submodules, or other data within any device. Thus, the methods described herein may be encoded as executable instructions embodied within tangible non-temporary computer-readable medium, including, but not limited to, storage devices and / or memory devices. Such instructions, when executed by a processor, cause the processor to execute at least a portion of the methods described herein. Furthermore, as used herein, the term “non-temporary computer-readable medium” includes all tangible computer-readable medium, including, but not limited to, volatile and non-volatile media, as well as removable and non-removable media such as firmware, physical and virtual storage, CD-ROMs, DVDs, and any other digital sources such as networks or the Internet, and digital means not yet developed, with the sole exception being temporary propagating signals.

[0020] Embodiments of the present invention generally relate to systems and methods for analyzing the physical characteristics of adsorbents for post-combustion carbon capture ("PCC"), and more particularly to systems and methods for modeling aspects of metal-organic frameworks ("MOFs") based on their components and their performance in carbon dioxide (CO2) capture capacity. In exemplary embodiments, the modeling system is configured to train a model based on the performance characteristics of known metal-organic frameworks ("MOFs") and their associated components. This model may then be used to evaluate the expected performance of other MOFs based on their particular components. Such modeling can help scientists and engineers evaluate different MOFs without having to physically test each candidate, thus leading to the faster development of MOFs with improved performance in CO2 capture capacity.

[0021] Figure 1 shows an exemplary PCC modeling system 140 that can be used to predict how a particular MOF can function in capturing carbon dioxide (CO2) from exhaust gas 116 of a coal-fired power plant 110. In an exemplary embodiment, the power plant 110 generates electricity 112 which is distributed via a transmission and distribution grid 114. The power plant 110 also includes a post-combustion carbon capture (PCC) system 120 configured to remove CO2 from exhaust gas 116 before the treated exhaust is released into the environment. The exemplary PCC system 120 includes a metal-organic framework (MOF) 122 that performs the capture function of the PCC system 120. The MOF 122 consists of three main components, namely a metal 124, an organolinker 126, and one or more functional groups 128. During operation, the PCC system 120 yields several performance results, which are generally expressed as “capture performance” 130 (e.g., PCC capacity) for purposes described herein. The PCC modeling system 140 is configured to estimate how a particular MOF (e.g., a specific combination of components and various related physical features) will function (e.g., predict performance in a "real" setting) when implemented in the PCC system 120.

[0022] In an exemplary embodiment, the PCC modeling system 140 includes a data acquisition and preparation module 142 configured to identify and store data used to train a machine learning model (e.g., a supervised predictive model). The preparation module 142 stores historical records of various adsorbents (e.g., MOF122) and their associated characteristics and known performance. Each historical record includes, for example, adsorbent data such as the components of a particular adsorbent (e.g., information about the metal 124 used in the secondary construction unit i.e., SBU, information about the organolinker 126, and / or information about any functional group 128 or amine associated with a particular MOF122), various physical characteristics of the adsorbent (e.g., pore size, pore volume, pore size distribution, surface area, etc.), and known performance data about the adsorbent (e.g., isosteric heat of adsorption (Q) at a specific pressure and / or temperature). ST This may include, for example, Henry's Law constant, selectivity ratio, adsorbent capacity per gram of adsorbent in millimoles of CO2, and adsorbent productivity per gram of adsorbent in millimoles of CO2 per unit time. Each of these historical records may be stored as model data 160 (e.g., in a database) and may include other data described herein. Some of these records may be collected from published literature, and others may include data collected from laboratory measurement data or operational performance data of the PCC system 120 (e.g., data collected during operation).

[0023] The PCC modeling system 140 also includes an exploratory data analysis module 144 configured to help the analyst 102 perform univariate data analysis on aspects of historical data. In an exemplary embodiment, the exploratory data analysis module 144 is configured to generate a graph displaying a single component of a known MOF for a single feature, thereby allowing the analyst 102 to see the trend of the feature across various MOFs. For example, the analysis module 144 could generate plots of two types of MOFs, chemioadsorbents and physicoadsorbents, as well as how they perform (e.g., as components across those two types of MOFs) with respect to Henry's constant, selectivity ratio, heat of adsorption, PCC capacity, and PCC productivity. In another example, the analysis module 144 could generate plots of various metal types within a physicoadsorbent or chemioadsorbent MOF and their respective heats of adsorption (e.g., range of values, mean values, etc. for each metal type). These plots may be compiled and generated from historical records identified within the model data 160 and displayed to the analyst 102 via the display of the computing device 104. From this data, analyst 102 can further identify one or more secondary features, which may be combinations of two or more primary features, as will be further explained below. Such secondary features may be used in later model training and analysis.

[0024] In an exemplary embodiment, the PCC modeling system 140 also includes a correlation and bivariate analysis module 146 configured to analyze the model data 160 in terms of positive or negative correlations between various feature pairs of historical records. For example, the analysis module 146 can identify a set of primary features and possibly one or more secondary features and perform a correlation analysis between the relevant features. For each particular pairing of two features, the analysis module 146 calculates a correlation coefficient (e.g., across all records in the model data 160) that identifies how those two features correlate with each other (e.g., a positive coefficient indicates a positive correlation between the two features, and a negative coefficient indicates a negative correlation between those two features). These correlation coefficients could be, for example, Pearson correlation coefficients. These correlation coefficients may be displayed to the analyst 102 in matrix form, thereby allowing the analyst 102 to see the strength of the correlations between various feature pairs. In some examples, the matrix could be a heatmap colored along the spectrum based on the magnitude of the correlation coefficients, displaying each correlation coefficient in the colors of a color spectrum that have a strong positive correlation at one end of the spectrum (e.g., dark blue) and a strong negative correlation at the other end of the spectrum (e.g., dark red). In some examples, the analysis module 146 can generate a pair plot for each feature pair, thereby allowing the analyst 102 to evaluate what type of relationship the two features have (e.g., linear, nonlinear, etc.).

[0025] The PCC modeling system 140 also includes a regression analysis module 148 configured to perform regression analysis using model data 160. In exemplary embodiments, the regression analysis module 148 identifies one or more features of interest for the regression analysis. In some embodiments, an analyst 102 can examine the aforementioned correlation matrix and / or pair plots and select a subset of features of interest. In some embodiments, the regression analysis module 148 can automatically identify one or more features of interest for the regression analysis (e.g., based on the correlation coefficients of the matrix). For example, the regression analysis module 148 can select all features that have a positive correlation coefficient with PCC capacities above a certain threshold (e.g., a strong positive correlation). In some embodiments, the regression analysis module 148 can also select all features that have a negative correlation coefficient with PCC capacities below a certain threshold (e.g., a strong negative correlation). In some embodiments, the features used to initiate the regression analysis are manually selected by a user or the like.

[0026] Once a subset of features is identified, the regression analysis module 148 performs a regression analysis process starting with the feature subset. In some embodiments, a portion of the model data 160 may be automatically or manually identified as training data (e.g., used in this regression analysis process), and the remaining portion of the model data 160 (e.g., 60% training data, 40% test data) may be reserved to evaluate the resulting model. In exemplary embodiments, the regression analysis process is performed in one or more stages, each stage comprising performing a multiple linear regression analysis on the current feature set. Based on the results of each stage, one or more features may be removed for analysis in subsequent stages, leaving a further subset of the narrowed features for analysis in the next stage until a specific stopping criterion is met. In this final stage, the regression analysis module 148 generates a transfer function using the last remaining narrowed features (e.g., PCC capacity as a function of a specific narrowed feature). This transfer function can then be applied to a new MOF and its associated feature values ​​to predict how the new MOF may perform in terms of PCC capacity, productivity, etc.

[0027] Further details and related functions performed by the PCC modeling system 140 are described in more detail in relation to Figures 3 and 4.

[0028] Figure 2 is a graph 200 showing the decrease in CO2 adsorption productivity for an exemplary adsorbent (e.g., exemplary MOF) as it progresses from a synthetic powder measured under equilibrium conditions to a final film coating on a substrate measured under dynamic conditions. In the exemplary embodiment, the Y-axis 202 of graph 200 is the adsorbent productivity of the MOF in kilograms (kgCO2) of CO2 per kilogram (kg) of adsorbent per hour (hr). The X-axis shows several stages 204 of PCC productivity between equilibrium powder productivity, from the initial stage 204A through the dynamic time effect stage 204B, the film mass transfer effect stage 204C, the thermodynamic effect (working capacity between adsorption and desorption) stage 204D, to the final coating working productivity stage 204E. In this example, at the initial stage 204A, the adsorbent represents a productivity of approximately 0.58 kgCO2 / kg / hr in the synthetic powder measured under equilibrium conditions. At each of stages 204B-204E, the adsorbent encounters a variety of decreases in productivity based on the various effects described above. For example, in the dynamic time effect stage 204B, the adsorbent may encounter a production decrease of approximately 0.2 212, down to a total of approximately 0.35 kg CO2 / kg / hr. Furthermore, for example, in the film mass transfer effect stage 204C, the adsorbent may encounter a second production decrease of approximately 0.05 214, down to slightly below a total of approximately 0.3 kg CO2 / kg / hr. Furthermore, for example, in the thermodynamic effect work capacity stage 204D, the adsorbent may encounter a third production decrease of approximately 0.1 216, down to a final coating work productivity of approximately 0.2 kg CO2 / kg / hr 220. Therefore, in this example, the adsorbent may encounter a total production decrease of approximately 0.38 kg CO2 / kg / hr 222, or a decrease of approximately 65% ​​(e.g., a utilization rate of approximately 35%). This knockdown effect is specific to each adsorbent, film coating, and cycle time operation.

[0029] The decrease shown in the PCC productivity of the exemplary adsorbent of FIG. 2 generally depends on the coating type and measurement type characteristics, and can be attributed to a combination of multiple factors such as the dynamics of the adsorption cycle time where equilibrium cannot be achieved, the coating mass transfer resistance, and the heat of isosteric adsorption that heats the adsorbent and reduces the performance capacity. Without the PCC modeling system 140 of FIG. 1, all of these factors are generally determined experimentally to understand the ultimate productivity of a given adsorbent.

[0030] For initial adsorption at low concentrations, Henry's law constant determines the equilibrium capacity of the adsorbent. The adsorbent capacity N i of adsorbent i is i =H i C i as in, the gas phase concentration C i of the adsorbent multiplied by Henry's law constant H i . The larger the value of Henry's constant, the steeper the initial slope of the isotherm. This means that an adsorbent with a higher H i has a larger capacity for the adsorbate in this low-concentration linear region of the isotherm. Maximizing H i also increases the productivity of the adsorbent at low gas phase adsorbate concentrations. Therefore, the PCC modeling system 140 attempts to find an adsorbent with a high Henry's law constant for CO2 adsorption by analyzing how various MOF components vary independently with Henry's constant. Since Henry's constant is a thermodynamic term (e.g., not a kinetic term), the dynamics of the adsorbent may also be evaluated to determine the adsorbent's productivity. These measurements are usually not reported in the literature for known MOFs, so this PCC modeling system 140 can first focus on Henry's constant and identify the MOF component with the highest capacity for CO2 at the lowest gas phase concentration of CO2. If dynamic uptake data are available, they are used as data inputs to the PCC modeling system.

[0031] Figures 3A–3C are flowcharts illustrating an exemplary method for analyzing the expected performance of a promising adsorbent in a post-combustion carbon capture system. In some embodiments, method 300 may be performed by a PCC modeling system 140, and the identified adsorbent may be incorporated into the PCC system 120 in Figure 1. In an exemplary embodiment, method 300 includes, in operation 310, collecting or otherwise identifying model training data for a known adsorbent (e.g., an MOF for which known performance data exists).

[0032] The identified model training data, in exemplary embodiments, includes instances of individual adsorbents and their specific compounds, as well as known performance data stored in a database such as a model data database (e.g., various variables for a particular known adsorbent). Each adsorbent may include information about the components of its particular adsorbent instance (e.g., specific metals, organolinkers, and / or functional groups or functionalities), various physical characteristics of the adsorbent instance (e.g., pore size, pore volume, isostatic heat of adsorption, Henry's Law constant, selectivity ratio, Langmuir / BET area, uptake rate, etc.), known performance data for the adsorbent instance (e.g., PCC capacity and productivity), and possibly unique identifiers or other adsorbent data (e.g., classification as a physicoadsorbent or chemioadsorbent). Metals may include, for example, nickel, chromium, magnesium, copper, manganese, zirconium, zinc, cobalt, indium, iron, aluminum, dysprosium, titanium, potassium, or any combination or alloy.Examples of organolinkas include 1,4-dioxide-2,5-benzenedicarboxylate (DOBDC), 1,4-benzene-dicarboxylate (BDC), 4,4'-oxide-1,1'-biphenyl-3,3'-dicarboxylate (DOBPDC), 1,3,5-tri(1H-1,2,3trizole-4-yl)benzene (BTTri), 1,3,5-benzenetricarboxylate (BTC), 1H,5H-benzo(1,2-d:4,5-d')bistriazole (BBTA), 1,1'-biphenyl-4,4'-dicarboxylate (BPDC), and 1,1'-biphenyl-3,3',5,5'-tetracarboxylate. These can include boxylates (BPTC), 1,5-dioxide-2,6-naphthalenedicarboxylate (DONDC), 1,2,4,5-benzene-tetracarboxylate (BTEC), 1,4-bix(1H-pyrazole-4-ylethynyl)benzene (BPEP), 2,5-di(1H-1,2,4-triazole-1-yl)terephthalate (BTTA), 2,4,6-tris(3,5-dicarboxylphenylamino)-1,3,5-triazine (TDPAT), 2,3,5,6-tetrachloroterephthalate (TCDC), 4,4'-dibenzoic acid-2,2'sulfone (SBPDC), and others. Their functions can include, for example, open metal moieties [OMS], microporosity [MP], Lewis base moieties [LBS], polar functional moieties [PFS], and post-synthetic modifications [PSM]. While exemplary metals, linkers, functional groups, and physical characteristics are provided herein, it should be understood that others are also possible and within the scope of this disclosure. In some circumstances, such model training data for adsorbents may be collected and recorded manually (e.g., by analyst 102) within a model data database. In some embodiments, the PCC modeling system 140 may be configured to collect such data (e.g., from performance or test data captured via the PCC system 120, via other online databases).

[0033] In exemplary embodiments, this adsorbent data is used by the PCC modeling system 140 as model training data to train a machine learning model to analyze the expected performance of promising adsorbents (e.g., novel combinations of metals, linkers, and functional groups that have not yet been studied and tested under real-world conditions). This model training data can be used as labeled training data in model building (e.g., instances of model training data for specific "inputs" that have known results or "labels"). In some embodiments, one subset of the model training data may be identified for the purpose of training the model, and another subset may be identified for the purpose of testing the model (e.g., 60% of adsorbents are identified for training, 40% for testing, 70% for training, and 30% for testing). In some embodiments, the analyst 102 can manually identify adsorbents for use in training and testing (e.g., by specific rows in the database). In some embodiments, the PCC modeling system 140 can automatically identify the training subset and the test subset (e.g., using pre-configured proportions, randomly selected proportions, etc.).

[0034] In some embodiments, Method 300 also includes performing a univariate data analysis on the model training data in Operation 320. This data analysis includes evaluating specific individual adsorbent variables (e.g., specific components or features) against other variables (e.g., other components, features, or performance values) across the body of the training data. This analysis can be used to identify high-level trends for specific components or features.

[0035] In exemplary embodiments, the PCC modeling system 140 can provide an analyst 102 with a graphical user interface that displays plots of specific features against other target features of interest, thereby enabling the analyst 102 to investigate trends and relationships between those specific features. The interface can enable the analyst 102 to select primary variables (e.g., as domains) and quadratic variables (e.g., as ranges) of interest, and the PCC modeling system 140 can then calculate arithmetic mean / mean / median and / or ranges of values ​​for bar graphs, forest plots, etc. (e.g., for domain variables with discrete values ​​and continuous quadratic variables) or generate scatter plots, etc. (e.g., for continuous primary and quadratic variables) over all the training data having those variables. For example, the PCC modeling system 140 can generate plots of organic linker pairs of isosteric adsorption heat or Henry's law constant or mean pore size, and can separately identify physicoadsorbents or chemioadsorbents (e.g., through different colors or shading). In another example, the PCC modeling system 140 can generate plots of metals identified in the training data against isosteric adsorption heat or selectivity ratio.

[0036] In operation 330, in an exemplary embodiment, the PCC modeling system 140 generates a quadratic feature for use during model training. A quadratic feature is a combination of two or more primary features. The term "primary feature" refers to one of the known variables (e.g., existing main effect parameters for past adsorbents) such as isosteric heat of adsorption, Henry's law constant, pore diameter, pore volume, and surface area. The term "secondary feature" refers to a combination of two or more of those primary features, including the interaction between the two primary features (e.g., interaction parameters). In some embodiments, the analyst 102 can manually create a quadratic feature within the PCC modeling system 140 by specifying two or more primary features to be combined as a new quadratic feature of the model. In one example, the quadratic feature "A" is (BET area * pore volume (Vpore )) is created as, and the secondary feature "B" is (V pore *Isosteric heat of adsorption (Q st It is created as )) and the quadratic feature "C" is (BET area * Q st These secondary features may be created as follows. These secondary features may be used in the model training and analysis described below.

[0037] In some embodiments, the PCC modeling system 140 can select secondary features to use during model training by using an automatic selection process. For example, the PCC modeling system 140 can analyze the statistical significance of each primary feature through probability values, etc. A probability value (p-value) is a measure of the significance of a variable to the model. A p-value less than 0.05 means that the factor in the model is statistically significant with 95% confidence, which provides strong evidence against random results, and the probability of random results is less than 5%.

[0038] In some embodiments, the PCC modeling system 140 can select a set number of quadratic features. For example, if the number of primary features analyzed by the PCC modeling system 140 is n, the set number of quadratic features is n It can be C. In this example, (n+ n A total number of primary and secondary features equal to C) may be automatically selected by the PCC modeling system 140.

[0039] In operation 340, in an exemplary embodiment, the PCC modeling system 140 performs correlation and bivariate analysis of the model data. Referring here to Figure 3B, this analysis includes identifying a feature set for model training. In operation 342, a promising feature set is identified for model training. This feature set includes a list of primary features 302 (or primary features) and secondary features 304 (or secondary features) to be used for model training. As described above, the primary features 302 include a list of physical features or known variables provided to the adsorbent in the model data 160 (e.g., heat of adsorption, pore diameter, pore volume, Henry's constant, etc.), and the secondary features 304 include a combination of those features as defined in operation 330. The primary features 302 and secondary features 304 are collectively referred to as the “feature set” 306 for model training. In an exemplary embodiment, the primary features 302 include PCC volume, BET area, Langmuir surface area, pore volume (V pore ), isosteric heat of adsorption (Q st The secondary feature 304 includes the three exemplary secondary features described above, namely "A" = (BET area * pore volume (V) pore )), ``B''=(V pore *Isosteric heat of adsorption (Q st )), and "C" = (BET area * Q st ) is included. Therefore, the training data can be considered to define an n-dimensional space, where n is the number of features 306 used in model training.

[0040] In operation 344, the PCC modeling system 140 can generate pair plots or scatter plots for each unique combination of primary features 302 and secondary features 304 in the feature set 306, using a training data adsorbent selected from the model data 160. For example, for each unique combination of features 306, the PCC modeling system 140 can generate a first pair plot showing the BET area and PCC capacity across the training data, a second pair plot showing the Langmuir surface area and PCC capacity across the training data, and a V across the training data. poreThe system can generate a third pair plot showing the PCC capacity, and so on. The PCC modeling system 140 can display these plots within a graphical user interface for inspection and review by the analyst 102. In some combinations, these plots can help the analyst 102 identify whether there is a correlation between those two variables in the training data, whether that correlation is linear or nonlinear, and whether it is a positive or negative correlation. Furthermore, these plots can help the analyst 102 identify statistically significant variables in the training data.

[0041] In operation 346, the PCC modeling system 140 generates a correlation matrix 308 of correlation coefficients for various pairs of primary features 302 and secondary features 304 in the feature set 306. More specifically, in an exemplary embodiment, the correlation matrix 308 is an n × n matrix, where each unique feature 302, 304 in the feature set 306 is assigned both rows and columns (e.g., a square reflection matrix). Each cell in the matrix 308 represents any combination of two features 302, 304 in the feature set 306 (e.g., based on the specific row and column of that cell), and the value contained in that cell is a correlation coefficient representing the degree or magnitude of the correlation between those two particular features. For each combination of two features 302, 304, the PCC modeling system 140 calculates the correlation coefficient for those two features (e.g., across the training data) in operation 348. In exemplary embodiments, the correlation coefficients are normalized from a range between +1.0 and -1.0, where a more positive correlation between two features is closer to +1.0, a more negative correlation is closer to -1.0, and a neutral (e.g., weak or nonexistent) correlation is closer to 0.0 (e.g., Pearson's correlation coefficient). In operation 350, the PCC modeling system 140 populates specific cells with the correlation coefficients associated with those two features 302, 304. Once completed, this correlation matrix and associated values ​​may be displayed to the analyst 102 for inspection and consideration, and in some embodiments, may be presented as a heatmap (e.g., coloring each individual cell based on its value), thus allowing the analyst 102 to more easily see the positive and negative correlations between specific features. Further details regarding the exemplary correlation matrix 308 are provided in relation to Figure 4.

[0042] Returning to Figure 3A, exemplary method 300 continues to operation 360, where the PCC modeling system 140 performs a regression analysis to generate a transfer function that can help approximate the PCC performance of other (e.g., untested) adsorbents. More specifically, Figure 3C shows the exemplary regression analysis process of operation 360.

[0043] In exemplary embodiments, a set of model training data is identified from model data 160 for use during regression operation 360. In some embodiments, as described above, specific instances of the training data may be identified manually (e.g., by analyst 102) or selected (e.g., automatically) by the PCC modeling system 140. Thus, the training data represents which known adsorbents fall within the scope of this particular regression analysis, and the remaining model data may be used to validate the results of this training.

[0044] In operation 364, a set of features is selected for the initial feature set 380 (for example, from the primary features 302 and secondary features 304 of the complete feature set 306). In some embodiments, all features of the complete feature set 306 may be used initially as the initial feature set 380. In other embodiments, the analyst 102 may manually select features within the initial feature set 380 (for example, based on examining pair plots and / or correlation matrices). In yet another embodiment, the PCC modeling system 140 may automatically select the initial feature set 380. For example, the PCC modeling system 140 may selectively add all features 302, 304 to the initial feature set 380 that have correlation coefficients with PCC capacities that are above a predetermined threshold (e.g., greater than 0.2) or below a predetermined threshold (e.g., less than -0.2). These f features are then used as the current feature set 382 to begin the regression.

[0045] In operation 366, the PCC modeling system 140 performs multiple linear regression on the model training data using the current feature set (e.g., using a normal least-squares model fit of the training data) (PCC productivity is the dependent variable of interest, determined based on PCC capacity and CO2 uptake rate (e.g., adsorption dynamics)). Here, a linear regression method is applied to Equation 1 to study the dependence between the parameter of interest (e.g., PCC capacity, Å) and various physical attributes selected for the initial feature set 380 (e.g., the BET area, pore volume, isothermal energy, as well as the secondary features "A", "B", and "C" described above).

number

[0046] In this regression, each of the f current features 382 corresponds to the associated linear coefficient b. i Along with X i This is represented by Equation 1. Initially, all initial features 380 are included in the transfer function of Equation 1, and as the regression is iterated, the current features 382 with the highest p-value are removed until all current features 382 have p-values ​​below a predetermined value (e.g., less than 0.05 means they have a 95% probability of being statistically significant). Relevant variables in the model are variables that have probability values ​​("p-values") below a predetermined threshold.

[0047] More specifically, in an exemplary embodiment, at each step of the iteration, a p-value is generated for each feature remaining in the current set of features 382. In test 368, the PCC modeling system 140 tests whether the regression is complete by evaluating the p-values ​​of the current features 382. If all the p-values ​​of the remaining current features 382 are below a predetermined threshold, the iteration terminates. Otherwise, the iteration proceeds to operation 370. In operation 370, the PCC modeling system 140 identifies the current feature with the highest p-value and removes that particular feature from the current features 382 for the next iteration. The regression process returns to operation 366 and continues with the reduced current set of features 382, ​​generating new p-values ​​again until all remaining features are below a predetermined threshold. If a primary feature has a p-value above the threshold, but a secondary feature containing the primary feature has a p-value below the threshold, the primary feature is retained as part of the model, even if its individual p-values ​​are greater than the threshold. This is necessary to retain secondary features that have p-values ​​below the threshold as part of the model.

[0048] Once the regression iterations are complete and finished, the PCC modeling system 140 may have identified one or more remaining features included in the final feature set 384, each of which has a p-value below a predetermined threshold and is therefore statistically significant. In an exemplary embodiment, the PCC modeling system 140 examines the final results for indications of potential overfitting in the model. For the final feature set 384, the PCC modeling system 140 also generates r-squared and adjusted r-squared values, and the difference between these two values ​​may indicate overfitting (e.g., if they are too far apart). If the difference between the r-squared and adjusted r-squared values ​​(e.g., abs(r-squared-adjusted r-squared)) is greater than a predetermined threshold, the model may be overfitted. In such a situation, the PCC modeling system 140 then analyzes each unique combination of the remaining features in the final feature set 384, runs the model on that combination to generate r-squared and adjusted r-squared values ​​for each combination, and then selects a particular combination that has the closest r-squared and adjusted r-squared values ​​(e.g., the smallest abs(r-squared-adjusted r-squared)).

[0049] The remaining features and their associated values ​​are used to generate the final transfer function from this model. More specifically, the regression yields the final coefficients for each of the remaining features in the final set of 384 features, as well as a constant coefficient (e.g., the y-intercept value). Thus, X in Equation 1 f Each of the variables is identified by each of the remaining features, and each association coefficient b f Furthermore, a constant coefficient b0 is added to Equation 1 to generate the final transfer function. This transfer function can then be used, along with prospective adsorbents and their associated values, to determine the expected PCC capacity of that particular adsorbent.

[0050] In some embodiments, the PCC modeling system 140 can use the model on test data to evaluate how it performs in terms of predictive ability. For example, the PCC modeling system 140 can generate residual plots of a linear regression model for both training and test data, thereby allowing an analyst 102 to evaluate how well the model performs.

[0051] In a specific example, the regression motion 360 is equivalent to the heat of isosteric adsorption (Q st ), BET area, pore volume (V pore This is performed on an exemplary training set starting with the initial feature set of three exemplary quadratic features "A", "B", and "C". In the first iteration, the BET area is identified as having the highest p-value of 0.948 and is removed. In the second iteration, the quadratic feature "A" is identified as having the highest p-value of 0.769 and is removed. In the third iteration, the quadratic feature "B" is identified as having the highest p-value of 0.441 and is removed. In the fourth iteration, all remaining current features (e.g., Q) are removed. st , V pore The p-values ​​for the "C" quadratic feature (and ) are below 0.05, and therefore the regression iterations terminate with these three features as the final feature set. However, in this example, the final r-squared value is 0.400 and the adjusted r-squared value is 0.363, and the difference triggers an overfitting analysis. Since there are three remaining features in the final feature set of 384, each combination of those features is unique to the model (e.g., a total of 3!=6 unique combinations, i.e., [Q st ], [Q st , V pore ], [Q st , V pore "C", [V pore ], [V pore It is checked in "C"] and ["C"]). In this example, [Q st The combination of ] alone yields r² of 0.567 and adjusted r² of 0.560 for a minimum difference of 0.007. Therefore, Q stOnly the features and their associated values ​​are used to generate a final transfer function that has only one remaining feature (e.g., X1=Q). st ), that is, Å = b1X1 + b0 = 0.0318*Q st -0.4243, where b1 = 0.0318, Q st The final coefficient generated for this is b0 = -0.4243, which is a constant coefficient. When the test data is applied to the trained model, in this particular example, a test r-squared value of 0.331 is obtained, and the r-squared value of the trained model was 0.567. Therefore, the formula includes only features for which the p-value is less than 0.05. Furthermore, a positive r-squared value in the test data indicates that the model has a heat of adsorption (Q st This indicates that approximately 60% of the capacity can be explained using ).

[0052] The PCC modeling system 140 can operate a post-combustion carbon capture system by using the generated transfer function to determine, but not limited to, one or more potential adsorbents to be used by the post-combustion carbon capture system. For example, the PCC modeling system 140 can identify one or more potential adsorbents based on carbon capture performance values ​​determined by the transfer function, in order to facilitate improvement of the overall carbon capture performance of the post-combustion carbon capture system.

[0053] Figure 4 shows an exemplary correlation matrix 308. In some embodiments, the correlation matrix 308 is generated by the PCC modeling system 140 and used in the method 300 described in Figures 3A–3C. In the exemplary embodiment, the correlation matrix 308 is a 9 × 9 square reflection matrix. There are 9 rows 402 and 9 columns 404, as well as 9 features 412, which are the subject of this exemplary model. Each of the 9 features 412 has both an associated row 402 and an associated column 404. The features 412 include 6 primary features 412A and 3 secondary features 412B, as in the example provided in Figures 3A–3C. Each cell in the matrix contains a correlation coefficient calculated between two specific intersecting features of that cell. For example, the correlation coefficient between BET area and PCC capacity is -0.12.

[0054] The methods, systems, and configurations disclosed herein are not limited to the specific embodiments described herein. Rather, steps of the methods, elements of the systems, and / or elements of the configurations may be used separately and independently of other steps and / or elements described herein. For example, the methods, systems, and configurations are not limited to the practice of the rotating machines described herein. Rather, the methods, systems, and configurations may be implemented and used in connection with many other applications.

[0055] Certain features of various embodiments are shown in some drawings and not in others, but this is simply for convenience. Furthermore, the reference to “one embodiment” in the above description should not be construed as excluding the existence of further embodiments that also incorporate the enumerated features. In accordance with the principles of this disclosure, any feature in the drawings may be referenced and / or claimed in combination with any feature in any other drawing.

[0056] This specification uses examples, including best modes, to enable a person skilled in the art to practice the disclosure, including the manufacture and use of any device or system and the execution of any incorporated method. The patentable scope of this disclosure is defined by the claims and may include other embodiments that a person skilled in the art may conceive. Such other embodiments are intended to be within the claims if they have structural elements that are not different from the language of the claims, or if they include equivalent structural elements that do not substantially differ from the language of the claims.

[0057] Further aspects of the present invention are provided by the subject matter of the following clauses.

[0058] Identifying a capture system for use in capturing carbon dioxide, a controller configured to operate the capture system, and a model training dataset, wherein each instance of the model training dataset identifies an adsorbent used by the capture system, a carbon capture performance value for the adsorbent, and a plurality of primary feature values ​​associated with a plurality of primary features; generating one or more secondary features based on one or more of the plurality of primary features, each of which is a combination of at least two of the plurality of primary features; determining a plurality of correlation intensity values, including correlation intensity values ​​between each of the plurality of primary features and each of the one or more secondary features; and determining a first of the model training dataset based on the plurality of correlation intensity values. A power generation system comprising a modeling system including a processor configured to identify a subset, determine statistical significance for each instance of a first subset of a model training dataset, identify a second subset of the model training dataset based on the statistical significance such that the statistical significance for each instance of the second subset of the model training dataset falls below a predetermined threshold, generate a transfer function based on the second subset of the model training dataset, and use the transfer function to determine one or more promising adsorbents to be used by the capture system based on carbon capture performance values, wherein a controller operates the capture system using one or more promising adsorbents determined by the modeling system.

[0059] A power generation system as described in any of the preceding clauses, which includes generating one or more secondary features based on one or more of a plurality of primary features, and performing univariate data analysis on a plurality of primary features.

[0060] A power generation system according to any of the preceding clauses, wherein determining multiple correlation intensity values ​​includes generating a correlation matrix containing multiple cells, wherein the multiple cells are arranged in multiple rows and multiple columns for multiple primary features and one or more secondary features, and each of the multiple cells contains one of the multiple correlation intensity values.

[0061] A power generation system as described in any of the preceding clauses, wherein the processor of the modeling system is configured to select an initial set of features from a plurality of primary features and one or more secondary features.

[0062] A power generation system as described in any of the preceding clauses, wherein the selection of an initial set of features includes receiving user input from a user indicating a selection of one or more features from a plurality of primary features and one or more secondary features.

[0063] The power generation system described in any of the preceding clauses, wherein selecting the initial set of features includes selecting one or more features from a plurality of primary features and one or more secondary features, each of which has a correlation coefficient greater than a predetermined threshold.

[0064] A power generation system according to any of the preceding clauses, wherein each of one or more features has one or more PCC performance data values ​​greater than a predetermined threshold.

[0065] The power generation system according to any of the preceding clauses, wherein the processor of the modeling system is configured to determine the r-squared value and the associated adjusted r-squared value for each unique combination of instances of a second subset of the model training dataset, determine the difference between the r-squared value and the associated adjusted r-squared value for each instance of the second subset of the model training dataset, and identify a third subset of the model training dataset based on the difference, wherein the difference for each instance of the third subset of the model training dataset is greater than a predetermined difference threshold.

[0066] A power generation system as described in any of the preceding clauses, wherein the per-instance difference value of a third subset of the model training dataset is less than a predetermined difference value threshold.

[0067] A power generation system according to any of the preceding clauses, comprising multiple primary features including at least carbon capture capacity and isosterically adsorbed heat generation.

[0068] A power generation system according to any of the preceding clauses, wherein one or more secondary features are based on one or more of the selectivity ratio, pore volume, and isosterically adsorbed heat generation.

[0069] A method for selecting one or more promising adsorbents to be used in operating a capture system to capture carbon dioxide, the method comprising: identifying a model training dataset, wherein each instance of the model training dataset identifies an adsorbent to be used by the capture system, a carbon capture performance value for the adsorbent, and a plurality of primary feature values ​​associated with a plurality of primary features; generating one or more secondary features based on one or more of the plurality of primary features, wherein each of the one or more secondary features is a combination of at least two of the plurality of primary features; determining a plurality of correlation intensity values, including correlation intensity values ​​between each of the plurality of primary features and each of the one or more secondary features; and based on the plurality of correlation intensity values A method comprising: identifying a first subset of a model training dataset; determining statistical significance for each instance of the first subset of the model training dataset; identifying a second subset of the model training dataset based on the statistical significance, wherein the statistical significance for each instance of the second subset of the model training dataset falls below a predetermined threshold; generating a transfer function based on the second subset of the model training dataset; and using the transfer function to determine one or more promising adsorbents to be used by the capture system based on carbon capture performance values, wherein the controller determines to operate the capture system using one or more promising adsorbents determined using the transfer function.

[0070] The method of any of the preceding clauses, which includes generating one or more secondary features based on one or more of a plurality of primary features, and performing univariate data analysis on a plurality of primary features.

[0071] The method according to any of the preceding clauses, wherein determining multiple correlation intensity values ​​comprises generating a correlation matrix containing multiple cells, wherein the multiple cells are arranged in multiple rows and multiple columns for multiple primary features and one or more secondary features, and each of the multiple cells contains one of the multiple correlation intensity values.

[0072] The method of any of the preceding clauses, further comprising selecting an initial set of features from multiple primary features and one or more secondary features.

[0073] The method of any of the preceding clauses, wherein selecting an initial set of features involves receiving user input from the user indicating a selection of one or more features from a plurality of primary features and one or more secondary features.

[0074] The method according to any of the preceding clauses, wherein selecting an initial set of features involves selecting one or more features from a plurality of primary features and one or more secondary features, each of which has a correlation coefficient greater than a predetermined threshold.

[0075] The method of any of the preceding clauses, further comprising determining the r-squared value and the associated adjusted r-squared value for each unique combination of instances of a second subset of the model training dataset, determining the difference between the r-squared value and the associated adjusted r-squared value for each instance of the second subset of the model training dataset, and identifying a third subset of the model training dataset based on the difference.

[0076] Identifying a third subset of the model training data includes determining that the per-instance difference value of the third subset of the model training dataset is greater than a predetermined difference value threshold, as described in any of the preceding clauses.

[0077] Identifying a third subset of the model training data is a method according to any of the preceding clauses, wherein the per-instance difference value of the third subset of the model training dataset is less than a predetermined difference value threshold.

[0078] Although the present invention is described in relation to various specific embodiments, those skilled in the art will recognize that the present invention can be put into practice with modifications within the spirit and scope of the claims. [Explanation of Symbols]

[0079] 102 Analyst 104 Computing Devices 110 Coal-fired power plants 112 Electricity 114 Power transmission and distribution networks 116 Exhaust gas 120 Post-combustion carbon capture (PCC) system 122 Metal-Organic Frameworks (MOFs) 124 Metal 126 Organic Linka 128 Functional group 130 Acquisition performance 140 PCC Modeling System 142 Data Acquisition and Preparation Modules 144 Exploratory Data Analysis Module 146 Correlation and Bivariate Analysis Module 148 Regression Analysis Module 160 Model Data 200 graphs 202 Y-axis 204 levels 204A Initial Stage 204B Exercise Time Effect Stages 204C Film material transfer effect step 204D Thermodynamic Effect Stage 204E Final Coating Work Productivity Stage 212 Production decrease 214 Production decrease 216 Production decrease 220 Final coating work productivity 222 Total production decline 300 ways 302 Primary Features 304 Secondary Features 306 Feature Set 308 Correlation Matrix 310 operation 320 operation 330 operation 340 operation 342 operation 344 operation 346 operation 348 operation 350 operations 360 operations 362 operations 364 operations 366 operation 368 operation 370 operation 372 operation 380 initial feature set 382 Current Feature Set 384 Final Feature Set 402 lines Column 404 412 Features 412A Primary Characteristics 412B Secondary Features

Claims

1. A capture system (120) for use in capturing carbon dioxide. A controller configured to operate the aforementioned capture system (120), and Identifying a model training dataset, wherein each instance of the model training dataset identifies an adsorbent used by the capture system (120), a carbon capture performance value for the adsorbent, and a plurality of primary feature values ​​associated with a plurality of primary features (302). The process involves generating one or more secondary features (304) based on one or more of the plurality of primary features (302), wherein each of the one or more secondary features (304) is a combination of at least two of the plurality of primary features (302). Determine a plurality of correlation intensity values, including the correlation intensity value between each of the plurality of primary features (302) and each of the one or more secondary features (304), Identifying a first subset of the model training dataset based on the plurality of correlation strength values, Determining statistical significance for each instance of the first subset of the model training dataset, Identifying a second subset of the model training dataset based on the aforementioned statistical significance, wherein the statistical significance of each instance of the second subset of the model training dataset falls below a predetermined threshold. To generate a transfer function based on the second subset of the model training dataset, Using the transfer function, determine one or more promising adsorbents to be used by the capture system (120) based on the carbon capture performance value. A modeling system (140) including a processor configured to perform the following: Equipped with, The controller operates the capture system (120) using the one or more promising adsorbents determined by the modeling system (140). Power generation system.

2. The power generation system according to claim 1, wherein generating one or more secondary features (304) based on one or more of the plurality of primary features (302) includes performing univariate data analysis on the plurality of primary features (302).

3. The power generation system according to claim 1, wherein determining the plurality of correlation intensity values ​​includes generating a correlation matrix comprising a plurality of cells, wherein the plurality of cells are arranged in a plurality of rows (402) and a plurality of columns (404) for the plurality of primary features (302) and the one or more secondary features (304), and each of the plurality of cells comprises one of the plurality of correlation intensity values.

4. The power generation system according to claim 1, wherein the processor of the modeling system (140) is configured to select an initial set of features from the plurality of primary features (302) and the one or more secondary features (304).

5. The power generation system according to claim 4, wherein selecting the initial set of features includes receiving user input from a user indicating the selection of one or more features from the plurality of primary features (302) and one or more secondary features (304).

6. The power generation system according to claim 4, wherein selecting the first set of features includes selecting one or more features from the plurality of primary features (302) and the one or more secondary features (304), each of which has a correlation coefficient greater than a predetermined threshold.

7. The power generation system according to claim 6, wherein each of the one or more features has one or more PCC performance data values ​​greater than a predetermined threshold.

8. The processor of the modeling system (140) Determining the r-squared value and the associated adjusted r-squared value for each unique combination of instances of the second subset of the model training dataset, Determine the difference between the r-squared value and the associated adjusted r-squared value for each instance of the second subset of the model training dataset, Identifying a third subset of the model training dataset based on the difference value, wherein the difference value for each instance of the third subset of the model training dataset is greater than a predetermined difference value threshold. The power generation system according to claim 1, configured to perform the following:

9. The power generation system according to claim 8, wherein the difference value for each instance of the third subset of the model training dataset is smaller than the predetermined difference value threshold.

10. The power generation system according to claim 1, wherein the plurality of primary features (302) include at least carbon capture capacity and isosterically adsorbed heat generation.

11. The power generation system according to claim 1, wherein the one or more secondary features (304) are based on one or more of selectivity ratio, pore volume, and isosterically adsorbed heat generation.

12. A method (300) for selecting one or more promising adsorbents to be used when operating a capture system (120) to capture carbon dioxide, wherein the method (300) is Identifying a model training dataset (310), wherein each instance of the model training dataset identifies an adsorbent used by the capture system (120), a carbon capture performance value for the adsorbent, and a plurality of primary feature values ​​associated with a plurality of primary features (302), Generating (330) one or more secondary features (304) based on one or more of the plurality of primary features (302), wherein each of the one or more secondary features (304) is a combination of at least two of the plurality of primary features (302), Determine a plurality of correlation intensity values, including the correlation intensity value between each of the plurality of primary features (302) and each of the one or more secondary features (304), Identifying a first subset of the model training dataset based on the plurality of correlation strength values, Determining statistical significance for each instance of the first subset of the model training dataset, Identifying a second subset of the model training dataset based on the aforementioned statistical significance, wherein the statistical significance of each instance of the second subset of the model training dataset falls below a predetermined threshold. To generate a transfer function based on the second subset of the model training dataset, The controller determines, using the transfer function, one or more potential adsorbents to be used by the capture system (120) based on the carbon capture performance value, and operates the capture system (120) using the one or more potential adsorbents determined by the transfer function. Method (300), including the method (300).

13. The method (300) of claim 12, wherein generating one or more secondary features (304) based on one or more of the plurality of primary features (302) includes performing univariate data analysis (320) on the plurality of primary features (302).

14. The method (300) of claim 12, wherein determining the plurality of correlation intensity values ​​comprises generating a correlation matrix (346) comprising a plurality of cells, wherein the plurality of cells are arranged in a plurality of rows (402) and a plurality of columns (404) for the plurality of primary features (302) and the one or more secondary features (304), and each of the plurality of cells comprises one of the plurality of correlation intensity values.

15. The method according to claim 12 (300), further comprising selecting an initial set of features (364) from the plurality of primary features (302) and the one or more secondary features (304).

16. The method (300) of claim 15, wherein selecting the first set of features (364) includes receiving user input from the user indicating a selection of one or more features from the plurality of primary features (302) and the one or more secondary features (304).

17. The method according to claim 15 (300), wherein selecting the first set of features (364) includes selecting one or more features from the plurality of primary features (302) and the one or more secondary features (304), each of which has a correlation coefficient greater than a predetermined threshold.

18. Determining the r-squared value and the associated adjusted r-squared value for each unique combination of instances of the second subset of the model training dataset, Determine the difference between the r-squared value and the associated adjusted r-squared value for each instance of the second subset of the model training dataset, Identifying a third subset of the model training dataset based on the aforementioned difference value. The method according to claim 12 (300).

19. The method according to claim 18 (300), wherein identifying the third subset of the model training data includes determining that the difference value for each instance of the third subset of the model training dataset is greater than a predetermined difference value threshold.

20. The method according to claim 18 (300), wherein identifying the third subset of the model training data includes determining that the difference value for each instance of the third subset of the model training dataset is less than a predetermined difference value threshold.

Images