Borehole induced structures seen in image logs from the Coso Geothermal Field, CA record variation in the azimuth of principal stress. Image logs of these structures from five boreholes were analyzed to quantify the stress heterogeneity for three geologically distinct locations: two boreholes within the Coso Geothermal Field (one in an actively produced volume), two on the margin of the Coso Geothermal Field and outside the production area, and a control borehole several tens of kilometers south of the Coso Geothermal Field. Average directions of Shmin and its standard deviation are similar along the eastern portion of the geothermal field at ~107 ± 28°; this is distinct from the western portion which has an azimuth of 081 ± 18° and also distinct from outside the geothermal field where the average azimuth is 092 ± 47°. These relationships suggest a correlation of stress orientation and heterogeneity with slip on the Coso Wash fault, suggesting that ~20 years of production has not affected the Shmin.orientation. The slope of power spectrum quantifies the length-scale dependence of stress rotations for the volume of the brittle crust penetrated by each borehole. Spectral analysis was applied to the depth variation of stress direction and it demonstrates that: (1) the data set contains distinct wavelengths of stress rotation, (2) that the relative power of these wavelengths in the total scaling of stress directions demonstrates a fractal distribution and (3) in a manner consistent with earthquakes causing the stress rotations. While the vertically averaged Shmin orientation for the three eastern boreholes varied by as little as 1°, the spectral slopes varied by 0.4 log (deg2 *m)(m) from the inside to the margin unproduced areas of the Coso Geothermal Field. The two boreholes inside the field had spectral slopes within one standard deviation, even though Shmin orientations were not parallel. These results suggest that at the kilometer length scale, the source of stress heterogeneity is dominated by proximity to recent fault slip, whereas the centimeter to meter stress heterogeneity is dominated by earthquake activity. / Geology

La connaissance des réseaux de fractures est essentielle pour comprendre la circulation des fluides dans un réservoir. Cette thèse s’appuie sur la reconnaissance du réseau de fractures naturelles qui chenalisent les circulations à l’échelle des deux forages profonds GRT-1 et GRT-2 de Rittershoffen (Alsace, France) qui ont recoupé les sédiments gréseux triasiques et le socle granitique altéré dans le cadre d’un projet industriel de géothermie. L’étude structurale de ce réseau de fractures a été réalisée à partir d’imageries de paroi acoustiques corrélées à des diagraphies géophysiques standard tandis que l’étude pétro-minéralogique se base sur les échantillons de cuttings. Les zones de fractures perméables des puits de Rittershoffen montrent une organisation asymétrique de la perméabilité. Des fractures ouvertes à l’échelle du puits semblent agir comme des drains perméables entourés de halos d’altération hydrothermale. Ces zones de fractures sont associées à des perturbations locales du profil de température dans le puits. La présence de minéraux illitiques hétérogènes pourrait être un indicateur pour prospecter les zones de circulations actuelles et passées à l’échelle des puits. Cette étude géologique permet d’évoluer vers un modèle de forage hydrothermal possédant des connexions favorables avec le réservoir sans avoir recours à des opérations de stimulation. / The knowledge of the fracture network is a key challenge to understand the fluid circulation through a reservoir. The aim of this PhD project is to investigate the natural fracture network that channelized the hydrothermal circulations into two deep wells GRT-1 and GRT-2 at Rittershoffen (Alsace, France) that intersect Triassic sandstones and altered granitic basement in the framework of an industrial geothermal project. The structural study of the fracture network was based on acoustic image logs correlated with standard geophysical logs, whereas the mineralogical study was based on cutting samples. Permeable fracture zones of wells at Rittershoffen present an asymmetrical organization of permeability. Open fractures at the borehole scale act as fluid pathways surrounded by halos of hydrothermal alteration. These fracture zones are associated with local thermal anomalies in the temperature profiles at the borehole scale. Occurrences of heterogeneous illitic minerals could be a good indicator to prospect zones of actual and past circulations at the borehole scale.

[pt] Este trabalho investiga o desempenho de modelos inteligentes na previsão de permeabilidade de reservatórios heterogêneos. Perfis de produção são utilizados para computar funções-objetivo para regressão no processo de otimização dos algoritmos. Um método de interpretação de perfil de produção é usado para remover efeitos de skin das medições de vazão. Adicionalmente, uma técnica de segmentação é aplicada a perfis de imagem acústica de alta resolução que fornecem, não apenas a imagem do sistema de mega e giga poros, mas também identifica fácies permeáveis ao longo do reservatório. A segmentação da imagem junto com outros perfis a poço aberto fornece os atributos necessários para o processo de treinamento do modelo. As estimativas apresentadas neste trabalho demonstram a habilidade dos algoritmos em aprender relações não lineares entre as variáveis geológicas e os dados dinâmicos de reservatório, mesmo quando a própria relação física é complexa e desconhecida à priori. Apesar das etapas de pré-processamento envolverem experiência em interpretação de dados, os algoritmos podem ser facilmente implementados em qualquer linguagem de programação, não assumindo qualquer premissa física de antemão. O procedimento proposto fornece curvas de permeabilidades mais acuradas que aquelas obtidas a partir de métodos convencionais que muitas vezes falham em prever a permeabilidade medida em testes de formação (TFR) realizados em reservatórios de dupla-porosidade. A contribuição deste trabalho é incorporar os dados dinâmicos oriundos dos perfis de produção (PP) ao processo de estimativa de permeabilidade usando algoritmos de Machine Learning. / [en] This work investigates the performance of intelligent models on the forecasting of permeability in heterogeneous reservoirs. Production logs are used to compute loss functions for regression in the algorithms optimization process. A flow profile interpretation method is used to remove wellbore skin effects from the measured flow rate. Additionally, a segmentation technique is applied to high-resolution ultrasonic image logs which provide not only the image of mega and giga pore systems but also identify the permeable facies along the reservoir. The image segmentation jointly with other borehole logs provides the necessary input data for the models training process. The estimations presented herein demonstrate the algorithms ability to learn non-linear relationships between geological input variables and a reservoir dynamic data even if the actual physical relationship is complex and not known a priori. Though the preprocessing stages of the procedure involve some data interpretation expertise, the algorithms can easily be coded in any programming language, requiring no assumptions on physics in advance. The proposed procedure provides more accurate permeability curves than those obtained from conventional methods, which may fail to predict the permeability measured on drill stem tests (DSTs) conducted in dual-porosity reservoirs. The novelty of this work is to incorporate dynamic production logging (PL) data into the permeability estimation workflow using machine learning algorithms.