O conhecimento de propriedades de transporte de misturas a diferentes pressões e temperaturas é importante em projetos, operação, controle e otimização de processos industriais. Nestes processos, frequentemente, o fluido é uma mistura binária ou multicomponente de hidrocarbonetos, como fluidos de petróleo. Propriedades experimentais de misturas, especialmente, a viscosidade absoluta como função de temperatura e pressão, podem fornecer importantes informações sobre o comportamento do fluido em diferentes composições e são usadas no desenvolvimento de modelos e correlações e na caracterização de misturas complexas. Diversas regras de mistura têm sido propostas na literatura para cálculo de viscosidade de misturas. Estas regras de mistura preveem o comportamento da mistura à pressão atmosférica usando propriedades dos componentes puros. Porém, em diversas aplicações é necessário estimar a viscosidade de misturas a altas pressões. Neste estudo, foram avaliadas regras de mistura comumente usadas como Refutas, Fator de Mistura, Índice de Mistura, Grunberg e Nissan, Kendall-Monroe e Eyring bem como Aditividade Molar, usando dados de viscosidade experimental de misturas em altas pressões. Inicialmente, foram realizadas medidas de viscosidade absoluta para a mistura altamente assimétrica de ciclohexano e n-hexadecano na faixa de temperatura entre (318,15 a 413,15) K e pressões até 62,053 MPa e, para este sistema, um modelo foi proposto para cálculo dos componentes puros para dada temperatura e pressão. Além disso, dados experimentais de viscosidade de trinta misturas cujos componentes diferem em forma, tamanho ou flexibilidade foram selecionados na literatura e modelados empregando-se regras de mistura. As viscosidades das misturas foram estimadas a partir de dados de viscosidade experimental dos componentes puros medidos nas mesmas temperaturas e pressões. A altas pressões, Refutas, Fator de Mistura e Índice de Mistura apresentaram os melhores resultados para todos os sistemas estudados. Mesmo para moléculas bastante assimétricas, Refutas, Fator de Mistura e Índice de Mistura podem ser usados. / The knowledge of transport properties of mixtures at different pressures and temperatures is important in design, operation, control and optimization of industrial process. In these processes, often, the fluid system is a binary or multicomponent mixture of hydrocarbons such as petroleum fluids. The experimental properties of mixtures, specially, the dynamic viscosity as a function the temperature and pressure, can provide valuable information about the fluid behavior at different compositions and are useful in developing models and correlations and in the characterization of complex mixtures. Several mixing rules have been proposed in the literature for calculating viscosity of mixtures. These mixing rules predict mixture behavior mainly at atmospheric pressure using pure component properties. However, in several important applications it is necessary to estimate the viscosity of blends at high pressures. In this work, were evaluate the performance of several commonly used mixing models like Refutas, Factor Mixing and Index Mixing, Grunberg and Nissan, Kendall-Monroe and Eyring as well as linear Molar Additivity, using experimental viscosity data of mixtures at high pressures. Initially, the absolute viscosities for the highly asymmetric mixture of cyclohexane and n-hexadecane were measured in the temperature range of (318.15 to 413.15) K and pressures up to 62.053 MPa and for this system, a satisfactorily model is proposed for calculating of pure components viscosities for given temperature and pressure. Also, viscosity data of thirty mixtures, whose components differ in molecular shape, size or flexibility, were selected and were modeled employing mixing. The mixture viscosities were estimated by all the mixing rules using experimental viscosity of pure components measured at the same temperature and pressure. At high pressures, Refutas, Factor Mixing and Index Mixing showed the best results for all systems studied. For very asymmetric systems, Refutas, Factor Mixing and Index Mixing can be used.

A global steady increase of energy consumption coupled with the decline of conventional oil resources points to a more aggressive exploitation of heavy oil. Heavy oil is a major source of energy in this century with a worldwide base reserve exceeding 2.5 trillion barrels. Management decisions and production strategies from thermal oil recovery processes are frequently based on reservoir simulation. A proper description of the physical properties, particularly oil viscosity, is essential in performing reliable modeling studies of fluid flow in the reservoir. We simulated cyclic steam injections on the highly viscous Hamaca oil, with a viscosity of over 10,000 cp at ambient temperature, and the production was drastically impacted by up to an order of magnitude when using improper mixing rules to describe the oil viscosity. This thesis demonstrates the importance of these mixing rules and alerts reservoir engineers to the significance of using different options simulators have built in their platforms to describe the viscosity of heavy oils. Log linear and power mixing rules do not provide enough flexibility to describe the viscosity of extra heavy oil with temperature. A recently implemented mixing rule in a commercial simulator has been studied providing satisfactory results. However, the methodology requires substantial interventions, and cannot be automatically updated. We provide guidelines to improve it and suggest more flexible mixing rules that could easily be implemented in commercial simulators. We also provide a methodology to determine the adequate time for each one of the periods in cyclic steam injection: injection, soaking and production. There is a lot of speculation in this matter and one of the objectives of this thesis is to better understand and provide guidelines to optimize oil production using proper lengths in each one of these periods. We have found that the production and injection periods should be similar in time length. Nevertheless, the production period should not be less than the injection period. On the other hand, the soaking period should be as short as possible because it is unproductive time in terms of field oil production for the well and therefore it translates into a negative cash flow for a company.

A global steady increase of energy consumption coupled with the decline of conventional oil resources points to a more aggressive exploitation of heavy oil. Heavy oil is a major source of energy in this century with a worldwide base reserve exceeding 2.5 trillion barrels. Management decisions and production strategies from thermal oil recovery processes are frequently based on reservoir simulation. A proper description of the physical properties, particularly oil viscosity, is essential in performing reliable modeling studies of fluid flow in the reservoir. We simulated cyclic steam injections on the highly viscous Hamaca oil, with a viscosity of over 10,000 cp at ambient temperature, and the production was drastically impacted by up to an order of magnitude when using improper mixing rules to describe the oil viscosity. This thesis demonstrates the importance of these mixing rules and alerts reservoir engineers to the significance of using different options simulators have built in their platforms to describe the viscosity of heavy oils. Log linear and power mixing rules do not provide enough flexibility to describe the viscosity of extra heavy oil with temperature. A recently implemented mixing rule in a commercial simulator has been studied providing satisfactory results. However, the methodology requires substantial interventions, and cannot be automatically updated. We provide guidelines to improve it and suggest more flexible mixing rules that could easily be implemented in commercial simulators. We also provide a methodology to determine the adequate time for each one of the periods in cyclic steam injection: injection, soaking and production. There is a lot of speculation in this matter and one of the objectives of this thesis is to better understand and provide guidelines to optimize oil production using proper lengths in each one of these periods. We have found that the production and injection periods should be similar in time length. Nevertheless, the production period should not be less than the injection period. On the other hand, the soaking period should be as short as possible because it is unproductive time in terms of field oil production for the well and therefore it translates into a negative cash flow for a company.

O conhecimento de propriedades de transporte de misturas a diferentes pressões e temperaturas é importante em projetos, operação, controle e otimização de processos industriais. Nestes processos, frequentemente, o fluido é uma mistura binária ou multicomponente de hidrocarbonetos, como fluidos de petróleo. Propriedades experimentais de misturas, especialmente, a viscosidade absoluta como função de temperatura e pressão, podem fornecer importantes informações sobre o comportamento do fluido em diferentes composições e são usadas no desenvolvimento de modelos e correlações e na caracterização de misturas complexas. Diversas regras de mistura têm sido propostas na literatura para cálculo de viscosidade de misturas. Estas regras de mistura preveem o comportamento da mistura à pressão atmosférica usando propriedades dos componentes puros. Porém, em diversas aplicações é necessário estimar a viscosidade de misturas a altas pressões. Neste estudo, foram avaliadas regras de mistura comumente usadas como Refutas, Fator de Mistura, Índice de Mistura, Grunberg e Nissan, Kendall-Monroe e Eyring bem como Aditividade Molar, usando dados de viscosidade experimental de misturas em altas pressões. Inicialmente, foram realizadas medidas de viscosidade absoluta para a mistura altamente assimétrica de ciclohexano e n-hexadecano na faixa de temperatura entre (318,15 a 413,15) K e pressões até 62,053 MPa e, para este sistema, um modelo foi proposto para cálculo dos componentes puros para dada temperatura e pressão. Além disso, dados experimentais de viscosidade de trinta misturas cujos componentes diferem em forma, tamanho ou flexibilidade foram selecionados na literatura e modelados empregando-se regras de mistura. As viscosidades das misturas foram estimadas a partir de dados de viscosidade experimental dos componentes puros medidos nas mesmas temperaturas e pressões. A altas pressões, Refutas, Fator de Mistura e Índice de Mistura apresentaram os melhores resultados para todos os sistemas estudados. Mesmo para moléculas bastante assimétricas, Refutas, Fator de Mistura e Índice de Mistura podem ser usados. / The knowledge of transport properties of mixtures at different pressures and temperatures is important in design, operation, control and optimization of industrial process. In these processes, often, the fluid system is a binary or multicomponent mixture of hydrocarbons such as petroleum fluids. The experimental properties of mixtures, specially, the dynamic viscosity as a function the temperature and pressure, can provide valuable information about the fluid behavior at different compositions and are useful in developing models and correlations and in the characterization of complex mixtures. Several mixing rules have been proposed in the literature for calculating viscosity of mixtures. These mixing rules predict mixture behavior mainly at atmospheric pressure using pure component properties. However, in several important applications it is necessary to estimate the viscosity of blends at high pressures. In this work, were evaluate the performance of several commonly used mixing models like Refutas, Factor Mixing and Index Mixing, Grunberg and Nissan, Kendall-Monroe and Eyring as well as linear Molar Additivity, using experimental viscosity data of mixtures at high pressures. Initially, the absolute viscosities for the highly asymmetric mixture of cyclohexane and n-hexadecane were measured in the temperature range of (318.15 to 413.15) K and pressures up to 62.053 MPa and for this system, a satisfactorily model is proposed for calculating of pure components viscosities for given temperature and pressure. Also, viscosity data of thirty mixtures, whose components differ in molecular shape, size or flexibility, were selected and were modeled employing mixing. The mixture viscosities were estimated by all the mixing rules using experimental viscosity of pure components measured at the same temperature and pressure. At high pressures, Refutas, Factor Mixing and Index Mixing showed the best results for all systems studied. For very asymmetric systems, Refutas, Factor Mixing and Index Mixing can be used.

Le développement du modèle prédictif E-PPR78, basé sur une méthode de contribution de groupe, a été entrepris depuis plus de dix ans pour prédire le comportement de systèmes multiconstituants. Ce modèle repose sur l'équation d'état de Peng-Robinson dans sa version de 1978 et les règles de mélanges de Van der Waals. Il utilise un seul paramètre d'interaction binaire, kij, qui dépend de la température. Afin de permettre au modèle E-PPR78 de prédire le comportement du gaz naturel, trois nouveaux groupes sont ajoutés : le monoxyde de carbone, l'hélium et l'argon. Pour cela, il a été nécessaire de former une base de données expérimentales la plus large possible contenant les mesures d'équilibres de phase et d'enthalpies de mélange pour les systèmes binaires constitués par ces trois groupes ainsi que ceux définis dans les études précédentes et présents dans le gaz naturel. Après une description de la classification des diagrammes de phase de Van Konynenburg et Scott, le modèle E-PPR78 est présenté. La troisième partie est consacrée à l'ajout des trois nouveaux groupes au sein du modèle. Les résultats sont obtenus avec une précision satisfaisante. Il apparaît clairement que le modèle E PPR78 est capable de prédire le comportement du gaz naturel dans des conditions de températures et de pressions particulièrement larges / The development of the predictive E-PPR78 model, based on a contribution group method, has been undertaken since ten years to predict accurately the behaviour of multi-component systems. This model lies on the Peng-Robinson equation of state with classical Van der Waals mixing rules. It uses a unique binary interaction parameter, kij, which is temperature dependant. To enable the E-PPR78 model to predict the behavior of natural gases, three new groups are added: carbon monoxide, helium and argon. It was necessary to build an experimental database, as exhaustive as possible, containing phase equilibrium and enthalpies of mixing data for binary systems formed by these groups and those defined in previous studies and present in natural gases. After a description of the classification scheme of Van Konynenburg and Scott, the E-PPR78 model is described. The third part is about the addition of the three new groups to the model. It clearly appears that the E-PPR78 model is able to predict the fluid-phase behavior of natural gases over wide ranges of temperatures and pressures

L’objectif principal de ce travail de thèse est de développer un modèle thermodynamique de type équation d’état cubique, permettant de prédire avec un maximum de précision les propriétés thermodynamiques des corps purs (des comportements de phases aux propriétés énergétiques - enthalpie, capacité calorifique - en incluant les propriétés volumiques) et des mélanges (équilibres de phases dans les régions sub- et supercritiques, points critiques, propriétés énergétiques, densités …), y compris les plus complexes. Concernant les corps purs tout d’abord : en nous appuyant sur la connaissance acquise par les études publiées pendant près d’un siècle et demi sur les équations d’état cubiques, nous avons identifié deux leviers pour accroître la précision de ces modèles. Le premier concerne la sélection d’une fonction α optimale (cette fonction est une quantité clef apparaissant dans le terme attractif du modèle) dont le bon paramétrage permet de représenter précisément les propriétés à saturation des corps purs, telles que la pression de saturation, l’enthalpie de vaporisation et la capacité calorifique du liquide à saturation. Afin que la fonction α puisse être extrapolée au domaine des hautes températures, nous avons défini les contraintes mathématiques que celle-ci doit respecter. Le second levier est le paramètre de translation volumique, paramètre clef pour la bonne représentation des densités liquides. Ces réflexions et les études associées sont à la base du développement des modèles tc-RK et tc-PR, utilisant une fonction α extrapolable à haute température ainsi qu’un paramètre de translation volumique, garantissant une précision jusqu’alors inégalée des propriétés sub- et supercritiques des corps purs prédites par des équations d’état cubiques. Afin d’étendre les modèles tc-RK et tc-PR aux mélanges, il a été nécessaire de développer des règles de mélange appropriées pour deux paramètres de l’équation d’état des mélanges : le covolume et le paramètre attractif. Des règles de mélanges récemment proposées qui combinent équation d’état et modèle de coefficient d’activité ont été adoptées. Les valeurs optimales des paramètres universels de ces règles de mélange ont été identifiées dans le cadre de cette thèse. Une règle de mélange linéaire pour le paramètre de translation volumique du mélange a été sélectionnée ; il a été prouvé que cette règle de mélange garantit l’invariance des propriétés d’équilibre de phases et des propriétés énergétiques entre les modèles translatés et non translatés. Afin de définir le modèle de coefficient d’activité optimal à intégrer dans la nouvelle règle de mélange, une base de données de 200 systèmes binaires a été développée. Ces systèmes binaires ont été sélectionnés afin d’être représentatifs des différents types d’interactions qui peuvent exister dans les mélanges non électrolytiques. La base de données accorde une place significative aux systèmes dits associés, qui sont certainement parmi les plus difficiles à modéliser par une équation d’état. In fine, cette thèse pose toutes les bases du développement d’une équation d’état cubique des mélanges. Le choix du modèle de coefficient d’activité optimal, la détermination des paramètres d’interactions binaires des 200 systèmes de la base de données et leur prédiction constituent des suites possibles de ce travail / The main objective of this thesis work is to develop a cubic equation of state thermodynamic model able to accurately predict the thermodynamic properties of pure compounds (from phase equilibrium data to energetic properties – enthalpy, heat capacity – and volume properties) and mixtures (phase equilibria in sub- and supercritical regions, critical points, energetic properties, densities…), including the most complex ones. Starting with pure compounds: relying on the knowledge collected all through the years from Van der Waals seminal work about cubic equations of state, we identified two levers to increase cubic-model accuracy. First is the selection of the optimal α function (this function is a key quantity involved in the model attractive term) the proper parameterization of which entails an accurate representation of pure-compound saturation properties such as saturation pressure, enthalpy of vaporization, saturated-liquid heat capacity. In order to safely extrapolate an α functions to the high temperature domain, we defined the mathematical constraints that it should satisfy. The second lever is the volume translation parameter, a key parameter for an accurate description of liquid densities. These studies led to the development of the tc-PR and tc-RK models, using an α function that correctly extrapolates to the high temperature domain so as a volume translation parameter, ensuring the most accurate estimations of pure-compound sub- and supercritical property from a cubic equation of state. In order to extend the tc-PR and tc-RK models to mixtures, it was necessary to develop adequate mixing rules for both equation of state parameters: the covolume and the attractive parameter. Recently proposed mixing rules combining an equation of state and an activity coefficient model have been retained. Optimal values of the mixing rules universal parameters have been identified in the framework of this thesis. A linear mixing rule for the volume translation parameter has been selected; it has been proven that this mixing rule does not change the phase equilibrium and energetic properties when switching from a translated to an untranslated model. In order to define the optimal activity coefficient model to include in the new mixing rule, a 200 binary-system database has been developed. These binary systems have been selected to be representative of the different kinds of interactions that can exist in non-electrolytic mixtures. The database includes in particular systems containing associating compounds, which are certainly among the most difficult ones to model with an equation of state. In fine, this thesis sets all the bases for the development of a cubic equation of state for mixtures. The selection of the optimal activity-coefficient model, the estimation of binary interaction parameters for the 200 binary systems from the database and their prediction are possible continuations of this work