Characterisation of chemical systems

专利摘要:
Disclosed is a method and system for characterising a target chemical system,´comprising. In use of the system, values of a physical property are obtained for a plurality of known chemical systems, a statistical moment of a charge density probability distribution of a known component of each of the known chemical systems is determined and the obtained values are fitted to a non-linear mathematical function of the statistical moment; to thereby determine an empirical relationship between the physical property and the statistical moment. This empirical relationship may be used to characterise a target chemical system, for example by predicting a value of physical property using the empirical relationship, or by "reverse engineering" to identify suitable chemical components.
公开号:DK201570244A1
申请号:DK201570244
申请日:2015-04-28
公开日:2015-05-04
发明作者:Kristian Mogensen；Martin Bennetzen
申请人:Mærsk Olie Og Gas As；
IPC主号:

专利说明:

Characterization of Chemical Systems
Field of the Invention
The invention relates to methods and apparatus for characterizing chemical systems, in particular the characterization of complex systems, for example in the field of enhanced oil recovery.
Background to the Invention
Characterization of chemical systems for industrial and research applications is most often based on a number of laboratory screening tests, to evaluate parameters such as compatibility of chemical components, thermal stability, rheology, phase behavior etc. For a given application it may be necessary to consider a number of properties of each component and this can be time consuming, or limit the number of properties or components that are considered.
In some applications, experimental work to identify whether a chemical or a formulation can be used in an application is prohibitively expensive or may not be technically feasible. One such application is the field of enhanced oil recovery (EOR).
In the primary stage of oil recovery, the natural pressure in the reservoir may be sufficient to drive the oil to the surface. However, as the reservoir is depleted, oil recovery enters a secondary stage in which water is injected into the reservoir in order to increase the pressure in the producing well. If the reservoir is further depleted, water flooding alone may not be sufficient to recover further oil. Various methods exist for recovering additional oil beyond this secondary stage, including chemical flooding, in which a fluid such as an aqueous solution to which surfactants or polymers have been added is injected into the reservoir to displace the oil.
Chemical additives change the balance of viscous and interfacial forces between the displaced fluid (oil) and the displacing fluid (water containing the chemicals) during a flood. This may be expressed in terms of the capillary number, where Ncap = µν / γ, and where µ is the viscosity of the displacing fluid, v the superficial velocity of the displacing fluid and the interfacial tension between the displaced and the displacing fluid.
The higher the capillary number obtained during a chemical flood the more oil is usually recovered, since interfacial forces are reduced on the microscopic scale. Ncap can be increased by increasing μ (e.g. by adding polymers to the displacing water) or by decreasing y (e.g. by adding surfactants to the displacing water).
Microscopic sweep efficiency depends not only on the capillary number but also on reservoir rock wettability, which reflects the thermodynamic preference of a solid to be in contact with one fluid rather than another, where both fluids are present. The displacing fluid may be capable of altering wettability from oil-wet to water-wet, hence releasing oii from the rock surface.
Selection of surfactants and polymers for a particular oil field application requires that a number of screening tests be performed for a number of physical parameters, which can be time consuming and costly.
In addition to the chemical systems encountered in the field of EOR, similar considerations may apply to types of chemical systems, such as certain biological systems, e.g. Determination of blood-brain barrier partition coefficients and intestinal adsorption of medical components or small biological molecules (K. Wichmann et al., "Prediction of Blood-Brain Partitioning and Human Serum Albumin Binding Based on COSMO-RS σ-Moments", J. Chem. Inf. Model., 47, 2007).
Hence, there is a significant need for methods of characterizing complex chemical systems which reduce the requirement to conduct experiments. However, chemicals and systems may be too complex to be described by simple thermodynamic models, for example due to inhomogeneity, lack of structural periodicity or complexity on various scales.
One method to predict fluid properties based on molecular structure is the UNIQUAC (UNIversai QUasiChemical) method (DS Abrams et al., "Statistical Thermodynamics of Liquid Mixtures: A New Expression for the Excess Gibbs Energy of Partly or Completely Miscible Systems", AiChE J., 21, 1975). UNIQUAC is an activity coefficient model used for phase equilibrium calculations. The model is a so-called lattice model and has been derived from a first order approximation of interacting molecule surfaces in statistical thermodynamics. However, UNIQUAC is not fully thermodynamically consistent because the local concentration around one central molecule is assumed to be independent of the local composition around another type of molecule. UNIQUAC is therefore typically useful only for describing phase equilibria of binary systems.
The UNIFAC semi-empirical (UNIQUAQ Functional-group Activity Coefficients) method is a development of UNIQUAC (e.g., Fredenslund et al., "Group-Contribution Estimation of Activity Coefficients in Nonideal Liquid Mixtures", AIChE J. 21, 1975). In the UNIFAC method, chemical structure of molecules is taken into account and molecules are subdivided into functional groups and an activity coefficient of a molecule in a liquid mixture can be calculated based on contributions from each of the functional groups present together with binary interaction coefficients related to the functional groups of other components of the mixture. However, as with the UNIQUAC theory which underpins it, the predictive capability of UNIFAC is limited for complex systems. A more thermodynamically consistent form of UNIQUAQ has been given by the more recent COSMO (the Conductor-like Screening MOdel) theory, as described for example by; A. Klamt et al., "COSMO-RS: a novel and efficient method for the a priori prediction of thermophysical data of liquids", Fluid Phase Equilibria, 172, 2000; A. Klamt, "The COSMO and COSMO-RS solvation models", Wiley Interdisciplinary Reviews: Computational Molecular Science, WIRE's Gomput Mol. Sci., 2011; and A. Klamt, "COSMO-RS: From Quantum Chemistry to Fluid Phase Thermodynamics and Drug Design", Elsevier Science Ltd., Amsterdam, The Netherlands, 2005.
Instead of using the functional group contribution method of UNIFAC, the main focus in COSMO theory is to characterize the screening charge density distribution on a molecular surface. For each kind of molecule X, a density functional calculation is performed in order to obtain the total energy Ex and the polarization (or screening) charge density σ of its molecular surface.
Charge density is a charge density and the local value of σ varies across a molecule. Hence, a surface charge landscape for a molecule can be represented by a charge density probability distribution. In COSMO, this distribution is called a 'sigma profile'. The sigma profile of an ensemble of molecules (eg a solvent) is calculated, as a weighted average of the mole fractions of the components in the ensemble, and integrated across all pairwise interactions to derive the "sigma potential profile" of the ensemble . The sigma potential is the energy associated with the preference for the solvent to interact with a surface element represented by charge density σ. The chemical potential of a molecule X within the ensemble may be characterized by its own sigma profile, and the profiles together can be used to calculate a variety of thermodynamic properties.
Both electrostatic any hydrogen bonding interactions may be taken into account in COSMO.
Although COSMO cannot be used to calculate the properties of complex systems from first principles, it has been shown that statistical moments (called "sigma moments") of the sigma profiles can be used to predict properties of more complex solvent-solute systems to a good approximation. Klamt and co-workers (Zissimos et al. "A comparison between the two genera of linear free energy descriptors of Abraham and Klamt", J. Chem. Inf. Comp. Sci., 2002) showed that sigma moments and so- called Abraham parameters can be used as free energy descriptors by a linear series, such as:
where c0 and c, · are constants, M, are sigma moments and μ is chemical potential. Thus, if a physical property Q can be described as a function of chemical potential f (u), then;
Sigma moments of fh order can be related to physical characteristics of a molecule. For example, the 0th sigma moment is the surface area of a molecule, the 1st sigma moment is the expectation value of the sigma profile, which equals the overall charge of a molecule, the 2nd sigma moment is a measure of the overall electrostatic polarity of a molecule and the 3rd sigma moment is a measure of the charge density asymmetry.
The approach of Zissmos et al. is to provide a physical interpretation of the relationship between Q and the statistical moments and their methods are consistently limited to the special case where properties can be related to functions of sigma moments raised to the first order only and where Qocf (p).
Consequently, there remains a need for improved methods of characterizing chemical systems.
Summary of the invention
According to a first aspect of the invention, there is provided a method of characterizing a target chemical system comprising: obtaining values of a physical property of a plurality of known chemical systems; determining a statistical moment of a charge density probability distribution of a known component of each of the known chemical systems; fitting the obtained values to a non-linear mathematical function of the statistical moment; to determine a relationship between the physical property and the statistical moment; and using the relationship to characterize the target chemical system.
The obtained values of the physical property may be determined empirically and for example may be measured values, or values of a physical property calculated or derived from experimental measurements of a chemical system,
The relationship may be a mathematical and / or empirical relationship.
References herein to "empirical" and "empirically" Include a posteriori, rather than solely a priori, methods; i.e. methods based on observation rather than theory. For example, a relationship may be based on an observed "best fit" between parameters (such as the physical property and the statistical moment), rather than being calculated or predicted based solely on a theoretical model.
As such, the relationship determined in accordance with the invention may be regarded as an empirical relationship between the physical property and the statistical moment. This is not to exclude that an empirical relationship may provide for some theoretical interpretation or insight, however.
Similarly, an empirical value of a physical property may be based at least in part on an experimental observation. Moreover, the values of the parameters (e.g. the physical properties) upon which an empirical relationship is based may themselves be empirical values based on experimental measurements. An empirical relationship may be based (or additionally) on values calculated non-empirically. For example, an empirical relationship may be determined (for example, based on an observation of a correlation, trend, and / or fit) between two sets of values that are themselves determined theoretically, for example from a theoretical calculation.
An empirical value of a property need not be based on a direct experimental measurement of the property itself, but may be calculated based on one or more data obtained from experimental observation.
The method may comprise predicting a value of the physical property of the target chemical system, by: determining a corresponding statistical moment of the charge density probability distribution of a target component of the target chemical system, based on chemical structure information related to the target component ; and predicting the value of the physical property of the target chemical system from the statistical moment of the target component, using the empirical relationship.
The method may comprise evaluating the suitability of a target component for use in the target chemical system.
The suitability of the target component may be evaluated in any one of several ways. Suitability may be evaluated based on a comparison between a predicted value of one or more physical properties and a desired value of the or each physical property. The suitability of the target component may alternatively be evaluated based on a comparison of one or more parameters of the target component or the chemical system comprising it (which may be related to a predicted physical property or properties, such as statistical moments or a charge density probability distribution) and the corresponding parameter (s) which correspond to a desired value of the or each physical property.
According to the evaluated suitability may be a value of a physical property, such as the amount of that physical property by which two values differ, or may be a value related to it, such as a percentage by which the properties differ. Alternatively, the suitability may be a dimensionless value which may be used as a measure or comparator of suitability, e.g. for screening.
The method may comprise evaluating the suitability of a target component for use in the target chemical system, by comparing the predicted value of the physical property to a desired value of the physical property. A chemical system may comprise or consist of a chemical component. Accordingly, the target chemical system may comprise or consist of a target component.
That is to say, a chemical system may comprise a single component (e.g., a compound) or may comprise more than one component (e.g., a mixture of compounds, or a solution or interacting components of the same or a different phase). Thus, the physical property may be a property of a chemical component within a chemical system comprising multiple components (for example, the solubility of a compound in a solvent or solution, or a partition coefficient). The physical property may be a property of a chemical system (for example, a viscosity or surface tension of a mixture or solution). The physical property may be a property of a chemical compound, i.e. of a chemical system comprising a single chemical component.
The charge density of a chemical component (e.g., a compound), or a chemical component within a chemical system, is dependent on the local environment (e.g., the chemical structure) of the chemical component and may vary with location in relation to the chemical component. The charge density may be a molecular surface parameter. Accordingly, a charge density probability distribution (also known as a sigma profile) may reflect the probability of any point on a molecular surface (or, more generally, in the coordinate space of a chemical component) having a charge density value.
The statistical momentum of the charge density probability distribution may be a sigma moment. A non-linear mathematical function may be considered to be a mathematical function one or more with non-linear terms. A non-linear mathematical function may include both linear and non-linear terms. Non-linear terms may include, for example, log transformations, exponential functions, power functions. A linear mathematical function may be considered to be a zeroth or first order polynomial.
The "non-linear mathematical function of the statistical moment" may be a non-linear mathematical function having at least one non-linear term which includes the statistical moment not raised to the power 1.
The statistical moment of a charge density probability distribution of a known component of each of the known chemical systems may be determined based on chemical structure information relating to each known component.
For example, a sigma moment may be calculated using COSMO theory or a COSMO based method such as COSMO-RS.
The statistical moment of a charge density probability distribution of a known component of each of the known chemical systems may be determined based on another empirical and / or mathematical relationship. The other empirical relationship may be a relationship between the statistical moment and another physical property of the known component or chemical system, and / or of another chemical system comprising the known component. For example, empirical relationships have been shown between sigma moments and Abraham parameters (Zissimos et al., J. Chem. Inf. Comp. Sci. (2002), 1320-1331).
Thus, the method may comprise determining a first empirical relationship by fitting the obtained values to a non-linear mathematical function of a statistical moment which is determined based on a second empirical relationship.
The method may comprise determining more than one statistical moment of a charge density probability distribution of a known component of each of the known chemical systems.
An / order statistical moment of a probability distribution p of a property x may be described as follows:
The method may comprise determining one or more statistical moments of / h order, where / is 0 or an integer greater than zero.
The method may comprise determining all of the statistical moments from 0th to / h order, where is / is 0 or an integer greater than 0. It will be understood that the method may be applied to any number or selection of statistical moments of the probability distribution. In practice, / is typically in the range of 0 to 10, or from 0 to 8.
The method may comprise determining values of one or more further descriptors of the known component. The method may comprise determining values one or more energies or energy descriptors, for example values related to dielectric properties and aromatic rings of the known component. The method may comprise determining a molecular weight and / or volume of the known component.
The non-linear mathematical function, and thus also the empirical relationship, may comprise temperature and pressure dependent terms.
The method may comprise fitting the obtained values Qv of physical property v to nonlinear mathematical function given by equation (1):
Equation (1) where: coefficients a, b "ch drd5, /, j, k ·, ki-ks, l" g, and qrQs are real numbers; fs g, h, u, r, s and w are functions;
Mi is the / * h order sigma-moment of the target component, given by;
where σ is charge density; V is molecular volume of the target component; tV is molecular weight of the target component;
Bring is the energy associated with aromatic ring systems of the target component;
Ediei is the dielectric energy of the target component;
Emqlecular is the molecular energy of the target component; / is 0 or an integer greater than 0; and v (P, T) denotes a function v of pressure P and temperature T. E ^ g may be zero for target components lacking an aromatic ring. Alternatively, d4 may be set to zero for target components lacking an aromatic ring, so the term is not included in calculations, σ may be related to electrostatic energy and optionally also hydrogen donor / acceptor energy.
Emqlecular may be calculated using a density functional algorithm, or a quantum mechanical algorithm (which may be a density functional algorithm), such as COSMO theory. Emqlecular may be the van der Waal energy.
The empirical relationship may be a non-linear relationship, or the empirical relationship may be a linear relationship.
Fitting may comprise removing outlying values from the obtained values. A set of obtained values may include outlying values of certain known components which are inconsistent with the obtained values of the other known components. For example, particular physical or chemical affinities within the chemical system, or structural or chemical differences to the other known components, may give rise to outlying values. Removal of outlying values may improve the quality of the fit of the obtained values to the non-linear mathematical function.
Removing outlying values may conceptually identify the influence of each obtained value on the quality of the fit of the obtained values to the empirical relationship (for example, the influence on an R2 value). Removing outlying values may conceptually identify the influence of each obtained value on coefficients of the empirical relationship (for example, the influence of each obtained value on values of a, b c,> d - d5, /, j, kj, ki-ks , 1., qi or qrqs).
Removing outlying values may comprise omitting an obtained value (eg performing the step of fitting, or repeating the step of fitting, based on obtained values from which outlying values have been removed) having an influence on the quality of fit or on the coefficients, as the case may be, above a threshold value.
Fitting may comprise applying one or more selection conditions, for example to identify outlying values. A selection condition may comprise an influence of a value obtained above a predetermined threshold.
The outlying values may be identified and / or removed by applying a selection condition.
The method may comprise determining one or more desired statistical moment (and, optionally, one or more further descriptors), from a desired value of the physical property, using the empirical relationship.
The method may comprise designing a target component for use in the target chemical system, using the one or more desired statistical moments. A chemical structure of the target component may be derived from the one or more desired statistical moments, using a COSMO based theory, for example, by identifying chemical structure information consistent with a charge density probability distribution or one or more statistical moments thereof.
The method may comprise evaluating the suitability of a target component for use in the target chemical system, by comparing the one or more desired statistical moments to the one or more corresponding statistical moments of a target component.
The target component may be a real component (e.g., an existing chemical compound) or may be a proposed component (e.g., a proposed compound having a proposed structure). Thus, the method may be used to design a target component, or a target chemical system, by determining that a target component gives rise to a suitable value of a physical parameter.
The suitability of a target component may be evaluated by determining a distance (such as a Euclidian or statistical distance) between two or more desired statistical moments and the corresponding statistical moments of the target component.
The method may comprise determining one or more further empirical relationships between one or more further physical properties and one or more statistical moments.
The method may comprise predicting a value of one or more further physical properties of the target chemical system.
The method may comprise determining one or more statistical moments (and, optionally, one or more further descriptors) from the desired values of the physical property and the one or more further physical properties, using the empirical relationships. The method may comprise evaluating the suitability of a target component, based on the desired values or the one or more statistical moments (and, optionally, one or more further descriptors) giving rise to them.
Where desired values of several (i.e., a set of two or more) physical properties are sought, the one or more statistical moments and / or further descriptors giving rise to each (as calculated from respective empirical relationships) may differ. Accordingly, the invention may comprise calculating one or more optimized statistical moments / descriptors, giving rise to a set of optimized values which are closest to the set of desired values.
The closeness of the sets of optimized and desired values may be evaluated by a measure of a Euclidian or statistical distance, between the optimized and desired values.
Accordingly, the method may comprise solving N simultaneous equations (corresponding to each of N empirical relationships to N physical properties), based on N desired values, for P optimized statistical moments (or, optionally, P optimized descriptors, including optimized statistical moments and one or more further descriptors).
In order to avoid overdetermination, N> P. Preferably, N> P
The mathematically optimal solution of such simultaneous equations may correspond to a value of one or more of the physical properties which differs by an unacceptable amount from the desired value, Accordingly, an acceptable optimized value may be equal to, or within a predetermined range of, the corresponding desired vaiue.
Accordingly, calculating the one or more optimized statistical moments (or descriptors) may comprise applying one or more exclusion conditions, such as to exclude solutions of the simultaneous equations corresponding to a value of a physical property which differs by more than a predetermined amount from the desired value.
An exclusion condition may be a difference of a predetermined percentage from the desired value. An exclusion condition may be a predetermined absolute difference from the desired value.
The method may comprise screening (for example, in silico) a library of target components (ie a group of two or more target components, and in some cases a plurality of target components), by evaluating the suitability of each target component in the library . Screening may determine which of the target components is most suitable (that is, identifying the target probability distribution having the highest value of suitability). Screening may comprise accepting target components with a suitability above a predetermined threshold and rejecting target components with a suitability below the predetermined threshold. Each of the screened target components may be ranked in order of their suitability. Screening may comprise ranking the target components in order of suitability.
The library may comprise information relating to each of the library's chemical components, including one or more or; values of one or more physical properties, chemical structure information relating to each of the chemical components; a charge density probability distribution, or one or more properties of a said probability distribution (e.g. one or more statistical moments). The library may comprise or be stored in a database and may comprise data concerning various chemical components. The data may be obtained using any suitable method, for example experimentally, theoretically and / or by extrapolation or interpolation from other experimental or theoretical data.
The library may comprise a group of two or more (or a plurality of) surfactants, polymers, proteins, enzymes, polysaccharides, amino acids and / or ionic liquids.
The one or more physical property or properties may each be selected from the group comprising: viscosity, interfacial tension, surface tension, contact angle (between a fluid component of the chemical system and a solid component of the chemical system), adsorption coefficient (e.g. to a particular rock or rock type present in a well), adsorption enthalpy, partition coefficient, diffusion coefficient, solubility (eg solubility in brine), partition coefficient, dielectric constant, rheological properties (such as a rate of change or other parameter descriptive of changes in viscosity with shear rate, temperature and / or concentration), thermal stability. The one or more physical property or properties may comprise at least one of viscosity, interfacial tension, surface tension, contact angle (between a fluid component of the chemical system and a solid component of the chemical system), adsorption coefficient (eg to a particular rock or rock type present in a well), adsorption enthalpy, partition coefficient, diffusion coefficient, solubility (eg solubility in brine), partition coefficient, dielectric constant, rheological properties (such as a rate of change or other parameter descriptive of changes in viscosity with shear rate, temperature and / or concentration), thermal stability.
The one or more physical property or properties may each be selected from the group comprising: viscosity, solubility and interfacial tension. The one or more physical property or properties may comprise viscosity and / or solubility and / or interfacial tension.
The method may comprise calculating a desired charge density probability distribution, based on the one or more desired statistical moments (for example using a generalized Lambda distribution method, as described for example in ZA Karian, "Fitting Statistical Distributions - The Generalized Lambda Distribution and Generalized Bootstrap Methods ", Chapman and Hall / CRC, 2000).
The suitability of a target component may be evaluated by comparing the desired charge density probability distribution with the charge density probability distribution of the target component.
Thus, the invention extends in a second aspect to a method of evaluating the suitability of a target component, comprising obtaining values of a physical property for a plurality of known chemical systems; determining at least one statistical moment of a charge density probability distribution of a known component of each of the known chemical systems; fitting the obtained values to a non-linear mathematical function of the statistical moment; to determine an empirical relationship between the physical property and each statistical moment: calculating one or more desired statistical moments, based on a desired value of the physical property, using the empirical relationship; calculating a desired charge density probability distribution, based on the or each desired statistical moment; and comparing the desired charge density probability distribution with the charge density probability distribution of the target component.
Comparing the desired charge density probability distribution with the charge density probability distribution of the target component may comprise evaluating the quality of the match between the probability distributions (for example, an average or integral of the distances between corresponding points of each of the probability distributions).
The charge density probability distribution of the target component may be calculated based on chemical structure information related to the target component, or may be calculated from one or more sigma moments determined based on a second empirical relationship.
The method may further comprise determining a corresponding statistical moment (or moments) of a target component of the target chemical system, and predicting a value of the physical property of the target chemical system from the statistical moment of the target component, using the empirical relationship .
The method may comprise determining one or more further empirical relationships between one or more further physical properties and one or more statistical moments.
The method may be used to design a chemical system, by designing a target component, or identifying a suitable target component (e.g. by evaluating the suitability of one or more target components of a chemical system), according to the method of the first or second aspect.
Accordingly, in a third aspect, the invention extends to a method of providing a chemical system (or one or more chemical components thereof), comprising designing a chemical system by a method in accordance with other aspects of the invention, and providing the designed or identified target component and, optionally, one or more further chemical components of the chemical system.
The method may be designing or providing a chemical formulation for use in enhanced oil recovery. The chemical formulation may comprise one or more chemical compounds for injection into a well, or one or more chemical compounds which may be added to a solution (for example brine) for injection into a well, so as to facilitate recovery of oil from the well. !!. The method may comprise providing the chemical formulation and injecting the chemical formulation into a well. The method may then comprise recovering oil from a well.
For example, the chemical formulation may increase the capillary number between oil in the well and the displacing fluid (compared to brine). The physical property may be, or the physical property may comprise one or more of; viscosity, interfacial tension and solubility. The chemical formulation may comprise one or more chemical components independently selected from the following groups of chemical components: wettability modifiers; Polymers; alcohols; basic (alkali) agents; acidic agents; gels, including water swellable gels; cross-linker molecules; surfactants; materials / compounds for fracking; salts; dissolved gases.
The method may comprise identifying a target component which, as mentioned above, may be an existing / known chemical component. The chemical system may be provided by obtaining an amount of the target component (eg by synthesizing the target component, in accordance with established synthetic methods known to those skilled in the art, or by purchasing an amount of the chemical from a chemical supplier) and mixing, bending or the like the target component with other components of the chemical system.
The invention is not concerned with any specific method for mixing or blending a chemical system comprising a formulation of various chemicals. It is well understood that such formulations may be provided in a variety of ways, according to specific requirements. For example, provision of a polymeric species within a solvent (e.g., aqueous) formulation may require pre-mixing with a co-solvent (e.g., an alcohol) or addition of a co-solvent in order to facilitate dissolution of the polymer. Dissolution may be aided by mechanical agitation, such as high-shear mixing, sonication and the like, and / or modification of the physical form of the polymer prior to dissolution e.g. by grinding so as to increase surface area.
Dissolution of components which are gaseous under, for example, ambient conditions may, for example, require modification of the temperature or pressure of a formulation, or use of apparatus in which the contact time of a gas with a liquid (e.g., an aqueous solvent or co. solvent formulation) is increased, for example using a cyclone pump or the like.
The target component, or the chemical system consisting of a target component, may be a new chemical. For example, the method may comprise determining a chemical structure which would give rise to one or more desired physical properties (either alone or within a chemical system). Accordingly, the chemical component / system may be synthesized alternatively or additionally, in accordance with established synthetic methods. For example, the target component may have a structure similar to a known chemical, and the method may comprise making one or more substitutions or additions of functional chemical groups of the known chemical.
In summary, therefore, the method may be used to characterize a chemical system in a number of ways, including predicting values of one or more physical properties of a chemical system and / or by determining one or more parameters of that chemical system related to a physical property or combination of properties. A chemical system may be characterized or further characterized by evaluating the suitability of one or more prospective components of respective chemical systems and / or by determining a composition, for example by designing a chemical system or by designing a particular chemical component. For example, a chemical system may be designed by identifying a suitable, or the most suitable chemical component out of a range of options, and / or a component may be designed having a structure giving rise to desired statistical moments or charge density probability distribution.
In a fourth aspect of the invention there is provided a system for characterizing a target chemical system comprising; a data store, operable to store values of a physical property of a plurality of known chemical systems; a processing resource, operable to; determine a statistical moment of a charge density probability distribution of a known component of each of the known chemical systems; to fit values obtained from the data store to a non-linear mathematical function of the statistical moment (e.g. received from a first processing module) and output an empirical relationship between the physical property and the statistical moment; and to characterize the target chemical system using the empirical relationship (e.g., output by a second processing module).
The processing resource may comprise one or more processing modules. The processing resource may comprise a first processing module, operable to determine a statistical moment of a charge density probability distribution of a known component of each of the known chemical systems; a second processing module, operable to determine a statistical moment of a charge density probability distribution of a known component of each of the known chemical systems, and; a characterization module, operable to characterize the target chemical system using the empirical relationship output by the second processing module.
The data store may comprise a data module, such as a data storage device or system (e.g. a hard drive). The data store may be operable to receive and temporarily store data, for example from a keyboard or from an external data source (e.g. over a network or from a data storage device).
The processing resource may be operable to determine the statistical moment (s) based on chemical structure information related to each known component.
Chemical structure information relating to each known component may be stored on and / or received by the data store.
The processing resource may be operable to determine the statistical moment (s) based on another empirical relationship.
The processing resource may be operable to determine more than one statistical moment of a charge density probability distribution of the known component of each of the known chemical systems. The processing resource may be operable to determine one or values of more further descriptors of the known component, such as values of one or more energies or energy descriptors, molecular weight and / or volume
Thus, the processing resource may further be operable to fit the obtained values to a non-linear mathematical function of each of the statistical moments received from the first processing module, and to output an empirical relationship between the physical property and the statistical moments (and , optionally, one or more further descriptors).
The processing resource may be operable to apply one or more selection conditions to the obtained values, to identify outlying values. The selection conditions may be applied by a selection module, such as an algorithm executed by the processing resource. A user of the system may manually remove (and in some embodiments also identify) outlying values, using the selection module, for example via a graphical interface. For example, the selection module may identify candidate outlying values, and a user may review the candidates and select which ones to remove. The selection module may identify and automatically remove outlying values.
The processing resource (e.g., the characterization module thereof) may be operable to predict a value of the physical property of the target chemical system from a corresponding statistical moment of the target component, using the empirical relationship. The corresponding statistical moment of the target component may be determined based on chemical structure information related to the target component. The corresponding statistical moment the target component may be determined by the first processing module, by the characterization module, or by a still further processing module.
The processing resource may be operable to design a target component for use in the target chemical system, using one or more desired statistical moments (and, optionally, one or more further descriptors). The one or more desired statistical moments, energies and / or energy descriptors may be calculated from a desired value of the physical property using the empirical relationship.
The processing resource may be operable to evaluate the suitability of the target component for use in the target chemical system, by comparing the one or more desired statistical moments to the one or more corresponding statistical moments of a target component.
The processing resource, or a characterization module, may be operable to screen a library of target components, by evaluating the suitability of each target component in the library. The library may be stored by the data store. The library may, for example, comprise chemical structure information relating to each of the target components in the library.
Whilst particular modules have been described herein, the functionality of one or more of those modules can be provided by a single module, processing resource or other component. Conversely, the functionality of a given module can be provided by two or more modules, processing resources or other components in combination. Reference to a single module encompasses multiple components providing the functionality of that module, whether or not such components are remote from one another, and reference to multiple modules encompasses a single component providing the functionality of those modules.
According to a fifth aspect of the invention, a computer readable medium comprising program code executable on a computing device (such as a system of the fourth aspect) is provided to perform methods in accordance with the invention.
Further preferred and optional features of each aspect of the invention correspond to preferred and optional features of any other aspect of the invention.
Description of the Drawings
Non-limiting embodiments of the invention will now be described with reference to the following drawings, in which:
Figure 1 shows a system for characterizing a chemical system, in accordance with the invention.
Figure 2 shows a work flow of the method by which the system of Figure 1 operates.
Figure 3 shows the relationships between values of water-component interfacial tension (Exp. IFT) and predicted values (Pred. IFT) as calculated using the empirical relationship of Equation (3).
Figure 4 shows the relationship between values of air / chemical component surface tension (Exp. Surface tension) and predicted values (Pred. Surface tension) as calculated using the empirical relationship of Equation (4).
Figure 5 shows the relationship between values of adsorption constants for alkane adsorption to ZSM-5 at 300 ° C (Exp. K) and predicted values (Pred. K) as calculated using the empirical relationship of Equation (5).
Figure 6 shows the relationship between values of adsorption enthalpies for molecules in the aqueous phase adsorption to activated carbon at 300 ° C (Exp. Log (H)) and predicted values (Pred. Log (H)) as calculated using the empirical relationship of
Equation (6).
Figure 7 shows the relationship between values of crysoscopic constants (Exp. Kf) and predicted values (Pred. Kf) as calculated using the empirical relationship of Equation (7).
Figure 8 shows the relationship between values of partition coefficients of a range of known components between DMPC lipid bilayers and water (Exp. LogK (w-DMPC) and predicted values {Pred. LogK (w-DMPC) as calculated using the empirical relationship of Equation (8) values predicted using COSMOtherm are also shown (underlined).
Figure 9 shows the relationship between values of dielectric constants coefficients of a range of known compounds (Exp. Dielectric) and predicted values (Pred. Dielectric) as calculated using the empirical relationship of Equation (9).
Figure 10 shows the relationship between values of viscosity of a range of known compounds (Exp. Viscosity) and predicted values (Pred. Viscosity) as calculated using the empirical relationship of Equation (10). Predicted values calculated following removal of outlying measured values are also shown.
Figure 11 shows the relationship between values of solubility of a range of known compounds in water (Exp. Log (S)) and predicted values (Pred. Iog (S)) as calculated using the empirical relationship of Equation (11).
Figure 12 shows the relationship between values of the octanol-water partition coefficient for chemical systems comprising a range of known compounds (Exp. Iog (Pow)) and predicted values (Pred. Iog (Pow)) as calculated using the empirical relationship of Equation (12). Predicted values of log (Pow) as calculated by COSMOtherm are also shown.
Figure 13 shows the relationship between Henry constants of a range of known compounds (Exp. Log (H)) and predicted values (Pred. Log (H)) as calculated using the empirical relationship of Equation (13).
Figure 14 shows the relationship between diffusion constants of a range of known compounds in water (Exp. Log (D. water)) and predicted values (Pred. Log (D. water)) as calculated using the empirical relationship of Equation (14). Predicted values calculated following removal of outlying values are also shown.
Figure 15 shows the relationship between diffusion constants of a range of known compounds in air (Exp. Log (D. air)) and predicted values (Pred. Log (D. air)) as calculated using the empirical relationship of Equation (15). .
Detailed Description of Example Embodiments
Figure 1 shows a system 1 for characterizing a target chemical system in accordance with the invention. The system includes a processing resource 2 and a data store 3. The processing resource includes first and second processing modules 5 and 7, and a characterization module 9. The system 1 further comprises an optional user interface 11.
Measured values of a physical property of a plurality of known chemical systems are stored by the data store 3. The first processing module 5 is operable to determine one or more statistical moments of a charge density probability distribution of a known component of each of the known chemical systems. The second processing module 7 is operabie to fit measured values obtained from the first data storage module to the non-linear mathematical function of Equation (2), which is described below. Using the empirical relationship determined by the second processing module, the characterization module 9 is operable to characterize the target chemical system and to output results to the user interface device 11. In the present embodiment, the processing modules 5 and 7 and the characterization module 9 are implemented in computing apparatus (the system 1), by means of a computer program having computer-readable instructions executable by a central processing unit (CPU) of the computing apparatus to perform the method of the embodiment. The system 1 may also include a hard drive and other components of a computing device including RAM, ROM, a data bus, an operating system including various device drivers, and hardware devices including a graphics card. Such components are not shown in the Figures for clarity.
Other embodiments of each unit may also be implemented in software, hardware or any suitable combination of hardware and software. For example, the various modules may be implemented as one or more ASICs (application specific integrated circuits) or FPGAs (field programmable gate arrays).
In use, the system 1 functions according to the generalized work flow set out in Figure 2, specific examples of which are described in further detail below. It will be appreciated by the skilled addressee that the sequence of steps may vary from the sequence set out in Figure 2 and described herein.
At stage 20, values of a physical property of a plurality of known chemical systems are obtained. For example, the values may be obtained by the first processing module 5 from the data module 3, and / or may be obtained by the data module from an external source, such as the user interface device, or over a network.
At stage 22, statistical moments of a charge density probability distribution of a known component are determined, for each of the known chemical systems. In the embodiment shown, the statistical moments are determined by the first processing module 5, using COSMO theory, based on chemical structure information related to the known components, stored on the data module 3.
At stage 24, an empirical relationship between the physical property and the statistical moments is determined by the second processing module 7, by fitting the measured values to the non-linear mathematical function of Equation (2).
At stage 26, the empirical relationship is used to characterize a target chemical system.
The system may comprise a computing device (such as a personal computer or a workstation), having a user interface (eg a keyboard and / or other user input device, one or more screens), one or more volatile and / or non-volatile data storage devices (eg a hard drive, RAM), functioning as a data store, and a processor.
Embodiments of the invention may be considered to incorporate a statistical pseudo-physical model for the characterization of chemical systems. Non-linear functions of one or more sigma moments, and optionally quantum mechanical energies, have been used to describe properties Qv of a chemical system, by an equation of the generalized form of Equation (2):
Equation (2) where: coefficients a. Bh ch drd5, i, j, kh krk5, 4 Φ and qrqs are reai numbers; f, g, h, u, r, s and w are functions;
Mi is the F order sigma moment of the target component, given by; M, - jρ (σ) σ'άσ where σ is charge density; V is molecular volume of the target component; W s molecular weight of the target component;
Ering is the energy associated with aromatic ring systems of the target component;
First, is the dielectric energy of the target component;
Eqosmo is the molecular energy of the target component; / is 0 or an integer greater than or equal to 0; and v (P, T) denotes a function v of pressure P and temperature T.
The functions f, g, h, u, r, s and w could be log transformations, exponential functions, power functions etc. ECo $ mo is the molecular energy calculated by COSMO theory. In alternative embodiments, this term may be calculated by another density function! theory algorithm or an algorithm based on quantum mechanics.
The non-linear functions have been used to describe macroscopic physical properties of a variety of chemical systems. The methods in accordance with the invention may be based only on statistical relationships between physical properties and statistical moments. It has been found that complex chemical systems are more accurately described by fitting values of a physical property to the non-linear functions, than has been previously possible.
Prediction of Physicai Properties
The system 1 was used to apply Equation (2) (described in detail, below) to 13 different chemical systems (Examples 1-13 set out below) to demonstrate its utility in predicting physical properties Qv of related (target) chemical systems, using the method described above. The method has been applied for prediction of the following properties: Interfacial tension, surface tension, adsorption coefficient and enthalpy, cryoscopic constants, lipid bilayer-water transfer partition coefficients, dielectric constants, viscosity, solubility, octanol-water transfer partition coefficients, Henry constants and diffusion constants. Data were retrieved from literature and COSMO calculations were performed using the COSMOthermX software (published by COSMO logic GmbH & Co. KG) to extract sigma moments.
The generalized statistical moments model outlined in Equation (2) (hereinafter referred to as the "GSM model") was applied, and multivariate statistical regression based on Pearson covariance matrix calculations was used to determine the correlation between the predicted adsorption energies and the sigma moment free energy descriptors of the molecules.
Statistical significance for each regression parameter was determined using statistics-statistics for P-value calculation. Statistical analysis was conducted using the statistical scripting language R.
Results are summarized in Table 1. In all cases, the model descriptors are associated with high statistical significance with correlation Revalues between 0.7-0.99 for completely different physical systems highlighting the generality and power of the GSM model. Revalues presented below are measures of how the GSM models explain the variance of the n data points presented based on correlation matrices of multidimensional nonlinear equations, and a high value of R2 is therefore indicative of a good predictive capability. Hence the GSM model formalism can predict properties of molecular systems with significant prediction power based on advanced multivariate statistics. As mentioned below, prediction power (as indicated by high R2 values) has been shown for physical properties of certain complex chemical systems, and for physical properties relevant to applications such as EOR.
Tabie 1 Significance codes for individual explanatory variables based on t-stanstics: <2E-16 '***', 0.001 0.01 0.05 0.1 '$ 1' Numbers in parenthesis () indicate
Revalues for correlations where outliers are removed. If sigma moments or other descriptors are transformed by e.g. a log or power function, this is indicated in the table. Abbreviations: IFT = Interfacial tension, SFT = Surface tension, KadS.Zeoiite = Adsorption constant for alkanes on zeolite, Hactc = Adsorption enthalpy of molecules to activated carbon, Kqmpc - Partition coefficient of molecules between water and DMPC-lipid bilayers, ε = dielectrics ! constant, η = Viscosity, S = Solubility, Pow = octanol-water partition coefficient, H - Henry constant, = Diffusion constant in water, Dajr "Diffusion constant in air. σ; = / <h order sigma moment
Details for each correlation summarized in Table 1 are as follows:
Example 1: Prediction of molecule / water interfacial tension (IFT)
The GSM equation (2) was applied to experimental values (/ 7 = 75, chemically diverse molecules) of known component / water IFT (of a set of known chemical systems comprising water and each known component), to find a non-linear empirical relationship between energy descriptors (sigma moments) and IFT. Results are shown in Figure 3.
The best fit empirical relationship between IFT was found by fitting the GSM equation (2) to the vaues of IFT, to be of the form of Equation (3).
Equation (3)
The quality of fit for! FT was R * ~ 0.96 based on correlation matrix calculations.
Literature values of I FT were also available for dodecane and hexadecane, however the fitting was performed on a set of I FT values not including those for dodecane and hexadecane.
Sigma moments σ0, σ2 and σ3 were calculated for each target component, using COSMOthermX, based on their chemical structures, and the empirical relationship of
Equation (3) was used to predict the I FT values for target chemical systems comprising target components dodecane and hexadecane. A comparison of the predicted and literature I FT values is shown in Table 2.
Table 2
A prediction accuracy for IFT of 10 and 3% was observed for dodecane and hexadecane, respectively.
Example 2: Prediction of component / air-interfacial tension (surface tension. SFT)
Experimental SFT data was obtained from http://www.dynesonline.com/visc_table.html. The GSM equation 2 was applied to experimental values (n = 62, chemically diverse molecules) of molecule / air surface tension (SFT) at 298.15K to find a non-linear relationship between energy descriptors and SFT. Results are shown in Figure 4.
The best fit empirical relationship between SFT was found by fitting the GSM equation (2) to the values of SFT, to be of the form of Equation (4).
Equation (4)
The quality of fit for SFT was R2-0.797 based on correlation matrix calculations.
Example 3: Prediction of constant adsorption for molecules adsorbed to Zeolite ZSM-5 at 3Q0oC.
Experimental data for adsorption constants of a range of known components to ZSM-5 were obtained from Arik et al. ‘High-temperature adsorption of n-alkanes on ZSM-5 zeolites: influence of the Si / AI ratio and the synthesis method on the low-coverage adsorption properties’, Microporous and Mesoporous Materials 60 (2003) pp. 111-124.
Equation (2) was applied to experimental values (/ 7 = 5. chemically similar molecules) of adsorption constants (Kads.zeonte) for alkanes adsorption to ZSM-5 at 300 ° C to find a nonlinear relationship between energy descriptors and Kads.zeolite .. Results are shown in Figure 5.
The best fit empirical relationship between Ksds "Ze0ijte was found to be in the form of Equation (5). Ads.zeolite = ^ + h ° + ^ EqUatlOO (5)
The quality of fit for Kadszeoiite was R2 ~ Q.9Q based on correlation matrix calculations.
Example 4: Prediction of adsorption enthalpy for molecules adsorbed to activated carbon at 300K.
Experimental data for adsorption enthalpies of a range of known components in the aqueous phase to activated carbon were obtained from C. Mehler, A. Klamt, and W. Peukert. 'Use of COSMO-RS for the prediction of Adsorption Equilibria', AlChE Journal. 2002, 43 (5).
Equation (2) was applied to experimental values (n = 20, chemically diverse molecules) of adsorption enthalpies (Hac / c) for molecules adsorbed to activated carbon in an aqueous phase at 300K to find an empirical relationship between energy descriptors and log (Hactc )}. Results are shown in Figure 6.
The best fit empirical relationship between adsorption enthalpy and sigma moments was found to be of the form of Equation (8). log (H) = ασ0 + Βσ2 + ca-. - + · d Equation (6)
The quality of fit for log (Hactc) was R * = 0.955 based on correlation matrix calculations.
Thus, the invention has been shown to provide high predictive capability for a complex chemical system comprising a range of components in water and a solid adsorbate.
Example 5: Prediction of crvoscooic constants.
Equation (2) was applied to experimental values (n = 6, chemically diverse molecules) of cryoscopic constants and (Kf) for molecules (ie for known chemical systems comprised of a single known component) to find an empirical relationship between energy descriptors and log (Kt). Results are shown in Figure 7.
The best fit empirical relationship between cryoscopic constants and sigma moments was found to be of the form of Equation (7). log (Kf) = ασ3 - + - bV2 + d Equation (7)
The quality of fit for log (Kf) was = 0.84 2 based on correlation matrix calculations.
Example 6: Prediction of partition coefficient for DMPC (dimvristov! Phosphatidyl choline) lipid bilaver-water transfer of molecules.
Experimental data for partition coefficients KDmpc of a range of known components between DMPC lipid bilayers and water were obtained from Patal et al., 'Physicochemical interpretation and prediction of the dimyrisoyl phosphatidyl choline-water partition coefficient', J. Mol. Struct. (Theochem) (2002), vol 593, pp. 9-18.
Equation (2) was applied to experimental values (n = 9, chemically diverse molecules) of Kdmpc to find a nonlinear relationship between energy descriptors and KDMPC. Results are shown in Figure 8.
The best fit empirical relationship between cryoscopic constants and sigma moments was found to be of the form of Equation (8). ag (KDMPC) - ασ0 + ba2 + ccr3 + d Equation (8)
The quality of fit for log {Kdmpc) was R2 = 0.918 based on correlation matrix calculations. Strikingly, the GSM model predicts DMPC-water partition coefficients with higher accuracy (R = 0.918) than the COSMOthermX thermodynamic model (Ff = 0.809), indicating that the method of the present invention provides for significantly improved prediction power.
Example 7: Prediction of dielectric constants of molecules.
Experimental data for the dielectric constants of a range of compounds was obtained from http://www.engineeringtoolbox.com/liquid-dielectric-constants-d__1263.html.
Equation (2) was applied to experimental values (n = 31, chemically diverse molecules) or dielectric constants for molecules (ε) to find a nonlinear relationship between energy descriptors and ε. Results are shown in Figure 9 and the best fit empirical relationship was found to be of the form of Equation (9). ε - ασ0σ22 + ba2 + aj42 + dEdiJ '+ / Equation (9)
The quality of fit for ε was R2 = 0.838 based on correlation matrix calculations.
Example 8: Prediction of viscosity at 298.15K of liquids consisting of single meals.
Equation (2) was applied to experimental values (n ~ 62, chemically diverse molecules) or viscosities (η) for liquids (each consisting of a single compound) to find a nonlinear relationship between energy descriptors and log (η). Results are shown in Figure 10 and the best fit empirical relationship was found to be of the form of Equation (10). logO ) - ασ + 6ct22 + ca52 + dV + / Equation (10)
The quality of fit for log (/ 7) was R2 ~ 0.701 based on correlation matrix calculations.
The correlation value is relatively low compared to the values obtained for other regressions presented here, but provides for some prediction power, since it corresponds to a non-squared Pearson correlation of R = 0.84.
Notably, by removing four outlying values (for glycerol, cyclohexanol, 2-aminoethanol and bromine) the quality of fit was improved to R2-0.7761 with the explanatory variables having the same individual regression significance values and the regression lines fully overlapping (hence only one regression lines is shown).
Example 9: Prediction of solubility.
Equation (2) was applied to experimental values (n = 23. Chemically diverse moieties) of solubilities (S) of a range of molecules in water, to find a nonlinear relationship between energy descriptors and S. Results are shown in Figure 11 and the best fit empirical relationship was found to be of the form of Equation (11). log (s) = ασ0 + baHBacc2 + ecrH3donl + d Equation (11)
The fit for iog (S) was R2 = 0.8491 based on correlation matrix calculations.
Example 10: Prediction of partition coefficient for octanol-water transfer of molecules.
Equation (2) was applied to experimental values (n = 23, chemically diverse molecules) of partition coefficients (Pow) for molecules to find a nonlinear relationship between energy descriptors and iog (Povv). Results are shown in Figure 12 and the best fit empirical relationship was found to be of the form of Equation (12). log ^) = ίΐσ0 + έσ2 + caHBacc2 + d Equation (12)
The fit for and {P0yV) was R2-0.9541 based on correlation matrix calculations. Predicted values of log (Pow) as calculated by COSMOtherm are also shown in Figure 12. Strictly, and as also noted for Example 6, the GSM model presented here is capable of characterizing octanol-water partition coefficients with much greater accuracy than the thermodynamic model COSMOthermX (R2 = 0.9113).
Example 11: Prediction of Henry constants of molecules
Equation (2) was applied to experimental values (n = 23, chemically diverse molecules) of Henry constants (/ - /) for molecules to find a nonlinear relationship between energy descriptors and log (H). Results are shown in Figure 13, and the best fit empirical relationship was found to be of the form of Equation (13). login) = ασ2 + ha HBacc2 + camdon2 + d Equation (13)
The fit for log (/ - /) was R2-0.9238 based on correlation matrix calculations.
Example 12: Prediction of diffusion constants of molecules in water
Equation (2) was applied to experimental values (n = 23 chemically diverse molecules) of diffusion constants (D ") for molecules in water to find a nonlinear relationship between energy descriptors and log (Dw). Results are shown in Figure 14, and the best fit empirical relationship was found to be of the form of Equation (14). log (A,) = ασ0 + ba21 + c Equation (14)
The fit for Sog (Dvv) was R2-Q, 7926 based on correlation matrix calculations. The correlation improved significantly (to R2 = 0.9047) if one outlying value (for o-Xylene in water) was removed.
Example 13: Prediction of diffusion constants of molecules in air
Equation (2) was applied to experimental values (n = 23 chemically diverse molecules) of diffusion constants (Da / r) for molecules in air to find an empirical relationship between energy descriptors and log (Dair). Results are shown in Figure 15, and the best fit empirical relationship was found to be of the form of Equation (15). - ασ2 + bW2 + c Equation (15)
The fit for log (Da / f) was F ^ = 0.852 based on correlation matrix calculations.
For each of Examples 1-13 above, coefficients a, b, c and d are real numbers and their exact values are not relevant to the present discussion and so are not quoted here.
The methods described herein have therefore been shown to be capable of predicting physical properties which may be important in the field of EOR to a high level of accuracy, such as viscosity, interfacial tension between an aqueous and an organic phase or adsorption properties. The methods therefore have utility in the design and provision of chemicals and chemical formulations for use in EOR applications.
For example, an empirical relationship (such as Equations (3) - (15)) could be used to investigate the predicted properties of chemical systems comprising one or more alternative chemical components - for example, to determine whether related chemical components (e.g., a polymer with a different average molecular weight) are predicted to improve a given physical property - without the need to prepare and test multiple chemical systems. The method can therefore identify the chemical system (or systems) predicted to have the most desirable properties, and a single chemical system, or a comparatively small number of systems can be provided for experimental evaluation. identification of Suitable Target Components from a Set of Desired Physical Properties
As described above, the predictive capability of the GSM model has been shown in relation to a range of individual physical properties. The GSM model may also be applied to the identification of target molecules of a target chemical system based on a set of desired physical properties. The system 1 was used to apply Equation (2) to calculate a set of optimized statistical moments corresponding to a set of desired values of physical properties.
In summary, an / 7-dimensional system of n non-linear equations, representing n physical pre-defined desired properties involving m property descriptors (ie dependent variables including sigma moments and, in the present examples, quantum mechanical energies), have been optimeret. Analytical and numeric algorithms to solve such multidimensional systems, and find the mathematically optimum solution have been shown. However, it has been found that a mathematically sub-optimal solution may in some cases lead to the identification of the most suitable target component.
Exclusion conditions have been applied to the solutions, in the form of specific thresholds within which predicted physical properties must fall, so that only solutions giving a predicted property Q, within a range of the corresponding desired property value Q * are allowed, and other solutions are excluded, ie the exclusion condition may take the form: q'-q -__- L <a where a is the predetermined threshold amount.
In alternative embodiments, other types of exclusion conditions can be applied alternatively, or in addition, for example Qi> b and Q, <c.
The current computational algorithmic implementation of the GSM model thus includes maximum freedom of flexibility for these exclusion conditions.
The following workflow has been applied to the determination of suitable target components: 1. n desired physical properties are set 2. n equations are defined and the n-dimenstonal system of non-linear equations of the form of Equation (2) is produced involving m property descriptors (including sigma moments and COSMO-derived quantum mechanical energies) and drdm. 3. k exclusion conditions are defined (e.g., based on accuracy, min / max value, etc.) 4. The optimum solution of the system of equations is identified analytically, including application of the exclusion conditions, to yield a mathematically suboptimal solution. 5. Euclidean distances D between the desired values of the physical properties {qi ° ..qn0} and the values {qi..qn) predicted from each set of possible solutions {di..dm} is calculated as:
8. Optimized properties S ^ d ^ -. Dm *), and thus the optimized solution, are determined based on the shortest Euclidean distance D *. 7. From the optimized solution the descriptors including the optimized sigma moments are identified. 8. From the optimum solution the properties {q-i * .. qn *} closest to the desired properties in the Euclidean n-dimensiona! space (see below) are identified. 9. Values of the descriptors di ~ dm are obtained (for example calculated using COSMO from chemical structure information) for target compounds in a database. 10. Based on (7) and (9), the target compounds are screened to find the best match (as evaluated by a measure of Euclidian distance in the d-dimensional space), so as to identify the most suitable compound.
In an alternative embodiment, statistical distances that take sample variance into account could be used instead of Euclidean distances, so
where s, 2 are sample variances for all predicted values for each property q, for each possible solution Sid ^ d ^).
Optionally, the underlying probability distributions corresponding to the sigma moments present in {drdm} and {di-dm *} can also be determined, using the Generalized Lambda Distribution (GLD) model. Thereafter, a similarity search can be conducted in order to evaluate the suitability of each of the compounds in the library.
The above workflow was implemented by a computer processor (such as processing resource 2) executing computer programming code in the language R.
An example set of 8 input desired properties and output optimized properties are shown in Table 3. In the example shown, exclusion conditions were based on percentage deviation of the optimized properties from the desired properties; as follows (abbreviations as per Tab 3): I FT error <15% logHc.act error <15% iogKDPMC error <20% ε error <40% iogS error <10% iogPow error <30% logH error <25% logDw error <10%
Table 3 Desired and Optimized Physical Properties
Abbreviations: I FT = intertacia tension with water, logH.Cact = binding energy to activated carbon, logK.dmpc = partition coefficient into an iipid bilayer, Epsiion = dielectric constant, logS = solubility in water, log Row = octanol / water partition coefficient , logH = Henry constant, logDw = diffusion constant in water.
The optimized values of Table 2 correspond to the optimized sigma moments, and the optimized value of the descriptor Eaieh set out in Table 4.
Table 4 Optimized Descriptors
The computation for identifying the optimized values of the physical properties and the descriptors required in the region of 3.15 billion calculations within the full solution space of 166 million possible combinations of each of the 8 physical properties and each of the 7 optimization parameters (the set of energy descriptors including the sigma moments and Edje!).
Any chemical having sigma profiles characterized by o0 ~ 130, σ2 ~ 20, σ3 = 10, σ4 = 100, crnB.a «> 2 = 0, σΗ8.αοη2 ~ 6 (where σΗΒ.β« 2 and anedons are sigma moments related to hydrogen bonding (donor and acceptor) properties) and Edier ~ 8 would be suitable candidates for having all the 8 desired properties specified and within the errors defined by the exclusion conditions. The chemical candidates are identified from the library based on shortest Euclidean distance calculations described above. Optionally, one or more conditional statements may be applied to the evaluation of the suitability of each of the molecules in the library, such as to exclude compounds with values of the sigma moments that differ by more than a predetermined threshold amount from each of the optimized values.
The optimized descriptors set out in Table 4 may then be compared against values of the corresponding descriptors (obtained using COSMOthermX from structure information readily available from chemical databases). The best match or matches (as evaluated by a measure of Euclidian distance between the descriptors) can therefore be expected to be associated with a desirable combination of physical properties. Thus, a potentially suitable chemical component can be identified, and chemical systems comprising that component provided, without the requirement to conduct experiments to investigate each of the physical properties of interest, or to obtain and potentially conduct multiple searches of data libraries.
Accordingly, the method may comprise characterizing a target chemical system, for example identifying a suitable target chemical component for use therewith; and obtaining or synthesizing (e.g., by conventional methods) an amount of that chemical component, together with an amount of any other components of the target chemical system. The various components may then be contacted (e.g. by blending, mixing and / or heating and the like) and the chemical system so provided may then be used as intended. For example, the chemical system may be used for enhanced oil recovery and its use may involve injecting the chemical system into a well. Oil, or an increased rate of production thereof, may then be recovered from the well.
While certain embodiments have been described, these embodiments have been presented by way of example only and are not intended to limit the scope of the invention. Indeed, the novel apparatus and methods described herein may be embodied in a variety of other forms. Furthermore, various omissions, substitutions and changes to the embodiments described herein may be made without departing from the scope and spirit of the invention.

权利要求:
Claims (26)
[1]
A method of characterizing a chemical target system, comprising: obtaining values for a physical property of a plurality of known chemical systems; determining a statistical momentum of a probability density distribution charge of a known component of each of the known chemical systems; adapting the obtained values to a non-linear mathematical function of the statistical momentum, thereby determining an empirical relationship between the physical property and the statistical momentum; and application of the empirical context to characterization of the chemical target system.
[2]
The method of claim 1, wherein the values of the physical property of the plurality of known chemical systems are determined empirically.
[3]
The method of claim 1 or claim 2, comprising the characterization of a chemical target system for use in enhanced oil recovery.
[4]
A method according to any one of the preceding claims, comprising obtaining the values of a physical property of a plurality of known chemical systems from a data store; and using a processing resource; - for carrying out the determination of the statistical momentum of a probability density distribution charge of a known component of each of the known chemical systems; - to perform the adjustment of the values obtained from the data store to a non-linear mathematical function of the statistical moments, thereby determining an empirical relationship between the physical property and the statistical momentum; and - to perform the characterization of the chemical target system using the empirical context.
[5]
A method according to any one of the preceding claims, comprising predicting a value for the physical property of the chemical target system by: determining a corresponding statistical moment of the probability distribution of charge density of a target component of the chemical target system, based on information on chemical structure relating to the target component; and predicting the value of the physical property of the chemical target system from the statistical momentum of the target component using the empirical context.
[6]
A method according to any one of the preceding claims, wherein the statistical momentum of a probability density distribution of a known component of each of the known chemical systems is determined based on chemical structure information relating to each known component.
[7]
The method of claim 6, wherein the statistical momentum of a probability density distribution of a known component of each of the known chemical systems is determined using a COSMO-based method.
[8]
A method according to any of the preceding claims, wherein the statistical momentum of a probability distribution of charge density of a known component of each of the known chemical systems is determined based on a different empirical context.
[9]
A method according to any one of the preceding claims, wherein the empirical relation is a non-linear relation.
[10]
A method according to any one of the preceding claims, wherein the method comprises removing external values from the obtained values.
[11]
A method according to any one of the preceding claims, wherein the method comprises determining more than one statistical moment of a probability density distribution charge of a known component of each of the known chemical systems.
[12]
A method according to any one of the preceding claims, comprising determining values for one or more additional descriptors of the known component, wherein the one or more additional descriptors are values of one or more of: one or more energies, a or more energy descriptors, a molecular weight of the known component, a molecular volume of the known component.
[13]
A method according to any one of the preceding claims, comprising adapting the obtained values Qv for physical property v to nonlinear mathematical function given by Equation (1):
<img img-format = "tif" img-content = "drawing" file = "DK201570244A1C00441.tif" id = "icf0001" />
Equation (1) where: the coefficients a, bt, c ,, drd5, i, j, k „k ^ ks, lir q, and qrq5 are real numbers; f, g, h, u, r, s and w are functions; Mi is the sigma moment of the target component given by:
<img img-format = "tif" img-content = "drawing" file = "DK201570244A1C00442.tif" id = "icf0002" />
where σ is charge density; V is the molecular volume of the target component; W is the molecular weight of the target component; Erlng is the energy associated with aromatic ring systems of the target component; Ediei is the dielectric energy of the target component; Emolecular is the molecular energy of the target component; / is 0 or an integer greater than 0; v (P, T) is a function v of pressure P and temperature Γ; and wherein one or both of E ng and d4 are zero for target components which do not have an aromatic ring.
[14]
A method according to any one of the preceding claims, comprising determining one or more additional empirical relationships between one or more additional physical properties and one or more statistical moments.
[15]
The method of claim 14, comprising calculating one or more optimized statistical moments, which yields a set of optimized values of a set of two or more physical properties which is nearly a set of desired values of the physical properties.
[16]
The method of claim 13, comprising solving N simultaneous equations corresponding to N empirical relationships to N physical properties, based on N desired values for P optimized statistical moments or P optimized descriptors, wherein the descriptors comprise optimized statistical moments and wherein N &gt; P.
[17]
A method according to any one of the preceding claims, wherein the physical feature (s) is selected from the group comprising: viscosity, interface voltage, surface tension, contact angle, adsorption coefficient, adsorption enthalpy, partition coefficient, diffusion coefficient, solubility, partition coefficient, dielectric constant, , thermal stability.
[18]
A method according to any one of the preceding claims, comprising assessing the suitability of a target component for use in the chemical target system by predicting a value for the physical property of the chemical target system from the statistical momentum of the target component using the empirical link; and comparing the predicted value with a desired value of the physical property.
[19]
A method according to any one of claims 1 to 17, comprising assessing the suitability of a target component for use in the chemical target system by: determining one or more desired statistical moments from a desired value for the physical property using the empirical link; calculating a desired probability density distribution, based on one or more desired statistical moments; and comparing the desired probability density distribution of charge density with the probability distribution of charge density for the target component; or comparing one or more desired statistical moments with the one or more corresponding statistical moments of a target component.
[20]
The method of claim 18 or 19, comprising screening a library of target components by assessing the suitability of each target component in the library.
[21]
The method of claim 20, wherein the library comprises a group of two or more of the following: surfactants, polymers, proteins, enzymes, polysaccharides, amino acids, ionic liquids.
[22]
The method of any one of claims 1 to 17, comprising: determining one or more desired statistical moments of a charge density distribution of a target component of the chemical target system from a desired value of the chemical target system using the empirical context; and designing a target component for use in the chemical target system using one or more desired statistical moments.
[23]
A method of providing a chemical system for use in enhanced oil recovery, comprising: identifying a suitable target component by the method of any one of claims 18 to 21; or designing a chemical target system by the method of claim 22; and providing the designed or identified target component.
[24]
The method of claim 23, comprising providing one or more additional chemical components of the chemical system.
[25]
The method of claim 23 or 24, comprising providing one or more chemical components independently selected from the following groups of chemical components: humidity modifiers; polymers; alcohols; basic (alkali) agents; acidic agents; gels, including water-swellable gels; crosslinking molecules; surfactants; materials / compounds for fracking; salts; dissolved gases.
[26]
A system for characterizing a chemical target system, comprising: a data store capable of operating to store values of a physical property of a plurality of known chemical systems; and a processing resource which may operate to: determine a statistical momentum of a probability distribution of charge density of a known component of each of the known chemical systems; adapt values obtained from the data store to a non-linear mathematical function of the statistical momentum and emit an empirical relationship between the physical property and the statistical momentum; and characterize the chemical target system using the empirical context.

类似技术:

---

公开号 | 公开日 | 专利标题

---

Ng et al.2015|Challenges and opportunities in computer-aided molecular design

---

Alkhatib et al.2020|Perspectives and guidelines on thermodynamic modelling of deep eutectic solvents

---

Mejia et al.2014|Use of equations of state and coarse grained simulations to complement experiments: describing the interfacial properties of carbon dioxide+ decane and carbon dioxide+ eicosane mixtures

---

Malde et al.2011|An automated force field topology builder | and repository: version 1.0

---

Llovell et al.2011|Modeling complex associating mixtures with [C n-mim][Tf2N] ionic liquids: predictions from the soft-SAFT equation

---

Nino-Amezquita et al.2010|Measurement and prediction of interfacial tension of binary mixtures

---

Zhang et al.2015|Molecular topology and local dynamics govern the viscosity of imidazolium-based ionic liquids

---

Gani2004|Computer-aided methods and tools for chemical product design

---

Liu et al.2000|Molecular electronegative distance vector | related to 15 properties of alkanes

---

Llovell et al.2013|Transport properties of mixtures by the soft-SAFT+ free-volume theory: application to mixtures of n-alkanes and hydrofluorocarbons

---

Venkatraman et al.2017|Predicting CO2 capture of ionic liquids using machine learning

---

Kurnia et al.2014|Evaluation of the conductor-like screening model for real solvents for the prediction of the water activity coefficient at infinite dilution in ionic liquids

---

Liu et al.2016|Using MD simulations to calculate how solvents modulate solubility

---

Hedegård et al.2013|Assessment of charge-transfer excitations with time-dependent, range-separated density functional theory based on long-range MP2 and multiconfigurational self-consistent field wave functions

---

Kang et al.2014|Algorithmic framework for quality assessment of phase equilibrium data

---

Diky et al.2009|ThermoData Engine |: software implementation of the dynamic data evaluation concept. 4. Chemical reactions

---

Diky et al.2012|ThermoData Engine | software implementation of the dynamic data evaluation concept. 7. Ternary mixtures

---

Raabe2016|Molecular simulation studies in hydrofluoroolefine | working fluids and their blends

---

Abooali et al.2014|Novel method for prediction of normal boiling point and enthalpy of vaporization at normal boiling point of pure refrigerants: A QSPR approach

---

Diky et al.2013|ThermoData Engine |: Software implementation of the dynamic data evaluation concept. 8. Properties of material streams and solvent design

---

Papadopoulos et al.2020|An approach for simultaneous computer-aided molecular design with holistic sustainability assessment: Application to phase-change CO2 capture solvents

---

Weiss et al.2016|GPCR-bench: a benchmarking set and practitioners’ guide for G protein-coupled receptor docking

---

Abedin et al.2017|Computational evaluation of mixtures of hydrofluorocarbons and deep eutectic solvents for absorption refrigeration systems

---

Noroozi et al.2017|Microscopic structure and solubility predictions of multifunctional solids in supercritical carbon dioxide: A molecular simulation study

---

Mastrogeorgopoulos et al.2017|Toward a simple predictive molecular thermodynamic model for bulk phases and interfaces

同族专利:

---

公开号 | 公开日

---

WO2015032506A1|2015-03-12|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

---

US8589130B2|2009-11-11|2013-11-19|Schlumberger Technology Corporation|Method of selecting additives for oil recovery|

---

法律状态:
2016-06-06| PHB| Application deemed withdrawn due to non-payment or other reasons|Effective date: 20160510 |

优先权:

---

申请号 | 申请日 | 专利标题

---

US201361874382P| true| 2013-09-06|2013-09-06|

---

EP13183418|2013-09-06|

---

EP13183418|2013-09-06|

---

US201361874382|2013-09-06|

---

EP2014002411|2014-09-05|

---

PCT/EP2014/002411|WO2015032506A1|2013-09-06|2014-09-05|Characterisation of chemical systems|

---