L’automatisation de la prise de décision dans des applications qui affectent directement la qualité de vie des individus grâce aux algorithmes de réseaux de neurones est devenue monnaie courante. Ce mémoire porte sur les enjeux d’équité individuelle qui surviennent lors de la vérification, de l’entraînement et de la prédiction des réseaux de neurones. Une approche populaire pour garantir l’équité consiste à traduire une notion d’équité en contraintes sur les paramètres du modèle. Néanmoins, cette approche ne garantit pas toujours des prédictions équitables des modèles de réseaux de neurones entraînés. Pour relever ce défi, nous avons développé une technique de post-traitement guidée par les contre-exemples afin de faire respecter des contraintes d’équité lors de la prédiction. Contrairement aux travaux antérieurs qui ne garantissent l’équité qu’aux points entourant les données de test ou d’entraînement, nous sommes en mesure de garantir l’équité sur tous les points du domaine. En outre, nous proposons une technique de prétraitement qui repose sur l’utilisation de l’équité comme biais inductif. Cette technique consiste à incorporer itérativement des contre-exemples plus équitables dans le processus d’apprentissage à travers la fonction de perte. Les techniques que nous avons développé ont été implémentées dans un outil appelé FETA. Une évaluation empirique sur des données réelles indique que FETA est non seulement capable de garantir l’équité au moment de la prédiction, mais aussi d’entraîner des modèles précis plus équitables. / Algorithmic decision-making driven by neural networks has become very prominent in applications that directly affect people’s quality of life. This paper focuses on the problem of ensuring individual fairness in neural network models during verification, training, and prediction. A popular approach for enforcing fairness is to translate a fairness notion into constraints over the parameters of the model. However, such a translation does not always guarantee fair predictions of the trained neural network model. To address this challenge, we develop a counterexample-guided post-processing technique to provably enforce fairness constraints at prediction time. Contrary to prior work that enforces fairness only on points around test or train data, we are able to enforce and guarantee fairness on all points in the domain. Additionally, we propose a counterexample guided loss as an in-processing technique to use fairness as an inductive bias by iteratively incorporating fairness counterexamples in the learning process. We have implemented these techniques in a tool called FETA. Empirical evaluation on real-world datasets indicates that FETA is not only able to guarantee fairness on-the-fly at prediction time but also is able to train accurate models exhibiting a much higher degree of individual fairness.

Pain perception can be drastically affected by external factors in the surrounding environment. Previous studies relying on subjective ratings of pain have shown that guided meditations supplemented by virtual reality (VR) can reduce the perception of pain more successfully than the audio alone. The research question for this study was: How does the efficacy of meditation as a means of pain mitigation differ with and without the application of VR? To answer this question, we used fNIRS to visualize changes in the dorsolateral prefrontal cortex (DLPFC), which has been shown to play a role in the larger pain matrix in the brain. To induce discomfort, we used a Cold Pressor Arm Wrap (CPAW), which is a variation of the more common Cold Pressor Task (CPT). This study was a within-subjects design where in each of the three conditions participants experienced CPAW. The three conditions were: (1) control (no VR & no audio), (2) audio (audio only), and (3) VR (audio + VR). We hypothesized that there would be a decrease in cortical activity in the DLPFC during CPAW when supplemented by VR meditation compared to audio-led meditation and no meditation, due to a decrease in perceived pain in the VR condition. Our results showed that VR meditation did significantly reduce perceived pain, according to the subjective pain ratings. Participants rated their discomfort the lowest in the VR condition, slightly higher in the audio condition, and highest for the control condition (in the absence of meditation). However, there were no significant differences in activity in the DLPFC between the conditions. The lack of any significant findings with the fNIRS data could be the result of many issues, including not having a large enough sample, physiological noise, improper localization of the pain ROIs, and not having a strong enough discomfort stimulus. The future directions for this study would seek to remedy these issues in order to obtain more informative fNIRS results. Overall, this study did show that VR meditation does effectively reduce pain perception and is therefore a promising tool for nonpharmacological pain management.

Due to the global warming and climate change, an increased frequency of high intensity rains and other disasters are expected all around the world. To predict this change in climate the IPCC has created a set of climate scenarios, RCPs, that will try to predict the future climate based on how much we are able to adapt and mitigate the effects we as a species have on the environment. This master thesis will seek to explore the possibility to use urban planning tools to help mitigate the increased effects and sizes of floods due to the global warming. To achieve this, a case study of Luleå is done, where urban indicators are used to improve the resilience of the city. This is combined with a policy study to see how Luleå compare to other cities in their policies that affect flood protection and mitigation. To have a strong flood protection system in a city several factors need to be considered. One of the biggest factors is what type of strategy is chosen. One possible strategy is resilience, this combines seeming paradoxes into a working flood protection and mitigation plan. Resilience improves the flood protection and mitigation by combining and improving the robustness, adaptability, and transformability of the city. This is done by, among other things promoting inter disciplinary cooperation, public cooperation and knowledge of flooding, and promoting the use of water in the city as an asset. Blue and green infrastructure could also be implemented into the city as these measures help improve the resilience of a city in many regards. Not just for flood protection, but it can also help mitigate the effects of droughts or heatwaves and improve the general wellbeing of the citizens. In the policy study it was found that different cities varied in both scale and strategy in their flood protection measures. All the cities that were looked at would also need to increase the scale of their protection and mitigation measures to mitigate the increased size and frequencies that the climate change brings. In Sweden, especially in the northern parts, the increased risk is not as high as in other parts of Europe. This is due to the land rise in Sweden mitigating the sea level rise. In Luleå the sea level and land rise are expected to fully mitigate each other until the year 2100. The policy study also showed that a history of flood related disasters did not necessary guarantee a strong flood protection scheme, but it would increase the probability of one. In the case of Luleå, the city has mostly focused on flood proofing buildings and infrastructure in the high-risk areas or those who are seen as critical to the society. Based on the analysis of the policies and indicators that were developed for Luleå, the city seems to have good protection from the current risks, such as a 100-year flood, flow, or rain. But the systems in place will most likely need to be expanded and developed further to mitigate the rising risk due to global warming. Some measures that can be implemented are related to the adaptability and transformability, like brochures that teaches the public about flood-protection and what to do and how to act in case of a large flood in the city.

Over the past several decades, flooding events in the United States have become the most frequent and costliest natural disaster. In the US, city and regional leaders are planning new water and flood mitigation infrastructure in response to the challenges of flooding, uneven urbanization, and racialized exclusion. Historically, projects to keep water out have never been universal or evenly applied. Yet, ‘learning to live’ with water, a key tagline in current sustainable development paradigms, masks how histories of racialized land development are entangled with contemporary water infrastructure projects and are productive of regional planning power. This dissertation centers racial capitalism in analysis of how contemporary water infrastructure projects are entangled with, and informed by, histories of racialized land development in the mid-Mississippi River Region. Through two case studies on flood mitigation infrastructure in eastern Missouri, I trace the historic development of infrastructures that shape the ongoing racialization of space, infrastructure (re)development and community vulnerability to flooding today. The case studies draw from a range of data, including archival research on histories of land and infrastructure development, participant observation of planning meetings, professional conferences, and local neighborhood initiatives, and field observations of the built environment. I argue that 1) scholarship concerned with social-environmental inequities should engage racial capitalism as a framework to “provincialize” urban theory and environmental racism as a means to theorize uneven infrastructural provisioning as a mode of urbanization that (re)produces social difference and value creation under racial capitalism, 2) the historical development of flood control in the Mississippi region was fundamental to the development of racial capitalism because it consolidated regional planning power through methods of social and environmental domination, and 3) contemporary infrastructural redevelopment and flood mitigation projects must contend with the path dependencies of structural racism to disrupt existing cycles of marginalization across social differences to deliver meaningfully on equity goals. Ultimately, this study finds that flood-mitigation infrastructures, including levees, floodways, and dams, on the Missouri River and gray and green stormwater infrastructure (GSI) in the City of St. Louis are embedded in broader social-environmental networks and regional power blocs, whose regional history and dynamics have created distinct patterns of uneven urbanization and vulnerability to flooding disasters. Because infrastructure projects are embedded in the built environment for decades, the social relations comprising their implementation, or lack thereof, reach into present and future development considerations. Thus, when planning projects fail to grapple with path dependencies of past infrastructure projects, they may reproduce structural racism and re-create patterns of uneven urbanization and vulnerability to flooding disasters. / Geography

Hydraulic fracture is a well-developed and widely used technology across petroleum upstream operations. The process involves the high-pressure injection of ‘fracking fluid’, mainly water containing proppants and thickening agents, into wells to form underground artificial fractures. The fracking fluid extends forward and supports those fractures to create diversion channels which lead to increased production. Therefore, it is of great significance to improve the permeability of the fractured formation, which leads to increased oil and gas production, and improved efficiency of oil and gas wells. Due to the importance of conductivity in such operations, the quality of s proppants are a critical factor affecting the fracturing efficiency and stimulation. For improvement of proppant, high strength often comes at the cost of increased density. During fracturing stimulation, it demands a higher flow rate, viscosities, and pumping pressure. It also results in the consumption of a significant amount of power, fresh water and produces more greenhouse gas and chemical pollution in the formation and the well site. This study investigated the production of an Ultra-Lightweight Proppant made by Graphene Aerogel (GA) and Epoxy Resin (ER) composite. GA with graphene as the basic structural unit is a low-density solid material with a high specific surface area, abundant nanoporous structure and good mechanical properties. Cured ER has the characteristics of small deformation and shrinkage, good dimensional stability, and high hardness. ER is one of the commonly used substrates in resin matrix composites. The main research contents of this study are as follows: 1. Using graphene oxide solution as a precursor to prepare graphene hydrogels, then make graphene aerogels through supercritical drying or freeze-drying. The graphene aerogel was placed in ER, followed by immersion in a vacuum environment, and then thermally cured to obtain a graphene aerogel-epoxy composite. The prepared GA has a network microstructure; the ER combined with the aerogel, and the composite material have good mechanical properties, strength is about 50 MPa. Specific gravity is 1.2, 55% lighter than silica sand and 63% lighter than ceramic proppant. 2. Spherical GA were produced using a droplet freezing method, which was combined with ER to create spherical fracturing proppants. The graphene aerogel-epoxy resin composite compressive strength is about 30 MPa, which is significantly higher than silica sand. The sphericity is about 0.9, better than silica sand (0.6-0.8) and same as ceramic proppant (0.9). its crushing rate is better than intermediate-density ceramic at 50, 70, and 100 MPa. The conductivity test showed that this new proppant was 30% and 50% higher than traditional silica sand and ceramic. 3. In the Field study, collecting core sampling and fluid analysis results to evaluate the porosity formation, permeability, fluid density and oil viscosity. Analyzation of the fracturing treatment data for the pressure testing data within the current field. Using the test results of the graphene aerogel spherical epoxy proppant (GAS-EP) modelling conducted for fracturing half-length and fracturing width, the result showing the new proppant can improve the fracturing length by 35% and increase oil Estimated ultimate recovery (EUR) volume about 10K bbl. 4. Based on field data and modelling results, horizontal well decline curve were generated for each scenario, using economic indicators, such as NPV value, payback time and others estimated that the break-even price of GAS-EP will be $2800/t. 5. Discuss the environmental benefits, using GAS-EP could reduce chemical additives by 4600L, reduce freshwater consumption 190m³ and reduce CO₂ emission 1.3 tons for a single fracturing stimulation. The study resulted in the production of a novel proppant with improved properties compared to the conventional materials. The results of this study provide a better understanding of the novel proppant properties and concluded that the novel proppant could safely use in the petroleum industry for enhance hydrocarbon recovery and significant environmental and economic benefits.

Reducing carbon emissions is a looming challenge that will be required to limit global warming. One approach is to replace energy from fossil fuels with renewable electricity that has low carbon footprints. The continuous decrease of renewable electricity prices makes electrochemical processes very promising for environmentally friendly production of fuels and chemicals. One of the mature electrochemical processes is hydrogen (H₂) production from water electrolysis. If only excess solar/wind electricity is used to power water electrolyzers to produce green H₂, the intermittency of the electricity supply will require low-cost electrolyzer technologies. Emerging membraneless water electrolyzers offer an attractive approach to lowering the cost of H₂ production by eliminating membranes or diaphragms that are used in conventional water electrolyzers. One aim of this dissertation is to understand the performance limits of membraneless water electrolyzers compared to the conventional designs. Another key electrochemical process to reduce carbon emissions is the conversion of carbon dioxide (CO₂) to value-added fuels and chemicals. CO₂ captured from air or flue gas needs to be extracted from carbon capture solution and pressurized before feeding to a conventional CO₂ electrolyzer. In order to avoid these energy intensive steps to make pressurized CO₂, there is a growing interest in developing membrane-based electrolyzers that can directly utilize the carbon capture solution to conduct electrochemical CO₂ conversion. This dissertation also explores a scalable membrane-free electrolyzer design that can convert carbon capture solution to syngas. In Chapter 2, a parallel plate membraneless electrolyzer is used as a model system to demonstrate a combined experimental and modeling approach to explore its performance limits. This modeling framework quantitatively describes the trade-offs between efficiency, current density, electrode size, and product purity. Central to this work is the use of in situ high-speed videography (HSV) to monitor the width of H₂ bubble plumes produced downstream of parallel plate electrodes as a function of current density, electrode separation distance, and the Reynolds number (Re) associated with flowing 0.5 M H₂SO₄ electrolyte. These measurements reveal that the HSV-derived dimensionless bubble plume width serves as an excellent descriptor for correlating the aforementioned operating conditions with H₂ crossover rates. These empirical relationships, combined with electrochemical engineering design principles, provide a valuable framework for exploring performance limits and guiding the design of optimized membraneless electrolyzers. This framework shows that the efficiencies and current densities of optimized parallel plate membraneless electrolyzers constrained to H₂ crossover rates of 1% can exceed those of conventional alkaline electrolyzers but are lower than the efficiencies and current densities achieved by zero-gap polymer electrolyte membrane (PEM) electrolyzers. Chapter 3 presents a packed bed membraneless electrolyzer (PBME) design for which liquid bicarbonate electrolyte flows sequentially through alternating porous flow-through anodes and cathodes. Within this design, hydrogen oxidation at porous anodes is used to produce protons that trigger in situ CO₂ release immediately upstream of porous cathodes, where electrochemical CO₂ reduction generates the desired product and returns the solution pH back towards its inlet value. By using the sequential flow-cell arrangement, the PBME offers the ability to mitigate large concentration overpotentials and non-uniform current distributions that naturally arise during scale-up of conventional membrane-based devices that rely on lateral flow of catholyte parallel to the surface of the electrodes. This study uses in situ colorimetric imaging to highlight the ability of PBME to rebalance pH across electrodes. In addition, results obtained with a multi-cell PBME “stack” demonstrate the scalability of this concept and reveal the ability to increase CO₂ utilization from 12.9% for a single-cell PBME up to 20.5% for a four-cell PBME operated under baseline conditions. Modeling results indicate that current utilization values >80% are theoretically possible for optimized multi-cell PBMEs operated at ambient pressure. Chapter 4 demonstrates a stacked PBME design and an elevated pressure system. Hydrogen oxidation at the anode and hydrogen evolution at the cathode are conducted in this device at pressures up to 5 atm. The pressure of the system can be held at a constant value by the use of a back pressure regulator, and product gases can be collected with a gas sampling bag that’s directly connected to the back pressure regulator. The stereolithography 3D printing technology is used to fabricate components of the PBME from clear resin, which is suitable for the elevated pressure operation. Post-processing of the components makes surfaces transparent for imaging bubble dynamics in the device. Within this system, HOR current density of 120 mA cm-2 can be achieved at different pressures. Finally, Chapter 5 provides concluding remarks and discusses future opportunities and challenges for membraneless electrolyzers for water electrolysis and electrochemical CO₂ conversion.

京都大学 / 新制・課程博士 / 博士(農学) / 甲第24244号 / 農博第2523号 / 新制||農||1094(附属図書館) / 学位論文||R4||N5415(農学部図書室) / 京都大学大学院農学研究科地域環境科学専攻 / (主査)教授 舟川 晋也, 教授 樋口 浩和, 真常 仁志 / 学位規則第4条第1項該当 / Doctor of Agricultural Science / Kyoto University / DFAM

This dissertation contains two collections of analyses, both broadly focused on energy system planning, but motivated by different research objectives in distinct geographic settings. Part I – Chapters I-III – evaluates decarbonization strategies in New York. These studies are characteristic of the primary energy-related challenge faced by the Global North: How can states cost-effectively meet time-bound emissions reduction targets? A series of linear programs are developed to answer this question, culminating in the System Electrification and Capacity TRansition (SECTR) model, a high-fidelity representation of the New York State energy system that characterizes statewide emissions and allows for comparative study of various decarbonization pathways. SECTR simulations indicate that prioritizing heating and vehicle electrification alongside an expansion of instate wind and solar generation capacity allows New York to meet recently legislated climate goals more affordably than through approaches that mandate substantial low-carbon electricity targets. Additional work also explores the optimal distribution of energy infrastructure within New York to meet specified decarbonization targets, along with the value of supply-side, demand-side, and bidirectional methods of system flexibility. Part II of this dissertation – Chapters IV-VII – is concerned with the energy system challenges faced by the lowest income countries. Set in the Ethiopian Highlands, this work first aims to locate smallholder irrigated areas, as irrigation has attendant energy requirements that are larger and more likely to generate supplementary sources of revenue compared to residential demands. Here, a novel classification methodology is developed to collect labeled data, train a machine learning-based irrigation detection model, and understand the spatial extent of model applicability. Across isolated plots of land as small as 30m by 30m, the resulting model achieves >95% prediction accuracy. Further studies then explore the system planning implications of simulated electricity demands associated with these irrigated areas.

Recent tsunami events have inflicted devastating damage to coastal communities. Existing design standards provide a certain level of evaluation of tsunami effects such that critical infrastructure can be designed to resist tsunamis. Tsunami momentum flux, used to design structures is a function of water level height and velocity of tsunami bores. Understanding tsunamis and developing mitigation measures is essential. So far, some mitigation measures have been suggested, and to improve them, further investigations are required. The design of tsunami inundation effects mitigation canals is one of the suggested solutions which has received limited attention. The first objective of this study was to investigate the effects of a rectangular canal on the hydrodynamics of turbulent bores before and after the canal by conducting a series of physical experiments. A dam-break wave was used to simulate the tsunami-like turbulent waves passing over a smooth and horizontal surface, in the presence and/or absence of a canal. Three canal water depths were used to model shallow, moderate, and deep conditions, and three canal widths were also selected to model narrow to wide conditions while the dam break waves were generated from three different impoundment depths in a reservoir located upstream of the canal. The dam-break wave propagation over a horizontal, dry, and smooth bed revealed four regimes describing the variations of bore height with time. The time to reach the maximum bore height and the quasi steady-state regime were correlated with each impoundment depth and an empirical formulation was proposed to estimate the onset of the quasi steady-state flow. The maximum bore heights measured before and after the mitigation canal location were approximately 40 % and 50 % respectively, higher compared with those recorded in the corresponding tests without the presence of a canal. The second objective of this study was to experimentally investigate the effects of canal depth on the time history of bore height and its velocity. The experimental results were used for calibration and validation of a developed numerical model. The rapid release of an upstream impoundment water depth was employed to generate a bore analogous to a tsunami-induced inundation. The time histories of wave heights and velocity were measured upstream and downstream of the canal. The recorded time-series of the water surface levels and velocities were compared with the simulation results and good agreement was found between experimental and numerical water surface profiles using a Root Mean Square Error (RMSE) and the Relative Error. Three turbulence models:, namely the standard k-ε, the Realizable k-ε, and the RNG k-ε were tested, and it was found that all turbulence models perform well but the standard k-ε model provided satisfactory accuracy. The velocity contour plots for shallow, medium, and deep mitigation canals showed the formation and evolution of jets of different characteristics. The energy dissipation and air bubble entrainment of the tsunami bore, as it plunged into a canal, increased as the canal depth increased, and the jet flow of the maximum bore velocity decreased with increased canal depth. It was found that the eye of the vortex in the canal moved steadily in the downstream direction. Generally, the bore fully plunged almost nearly into the middle of the canal and started to divide into two small vortices. The third objective of this study dealt with a sequence of numerical experiments conducted to investigate the impact of mitigation canals on the hydrodynamics of a tsunami-like turbulent bore moving across a flat bed. The effects of mitigation canal depth and its orientation on the reduction of maximum specific momentum and energy of turbulent bores crossing over it were investigated numerically. Variations in the ratio between the downstream and upstream maximum specific momentum and mean flow energy decreased as the canal depth increased, and the time history of the mean flow energy over a canal with a rectangular endwise profile revealed that the canal depth affects the jet stream of the maximum mean flow energy. As the canal depth increased, the period of time needed to dissipate the area of the jet stream with the maximum turbulent kinetic energy, vorticity, and energy dissipation rate decreased. Both the angled and perpendicular to flow direction canals caused the maximum specific momentum and energy of the turbulent bore to decrease downstream of the canal. The specific momentum and energy achieved their highest values for a canal orientation of 45º. The greatest reductions in maximum specific momentum for turbulent bores over canals with different depths and orientations were achieved for 𝜃 = 30°.

Due to the serious impact wind damage has on homes in the Gulf Coast region, policy makers, community developers, and homeowners are seeking ways to lessen impacts. One potential tool to increase properties’ resiliency in the event of a periodic and catastrophic event is wind mitigation, the process of adding features to a building, i.e. a house, to increase the strength of the structure amid a storm such as a hurricane. In this research, I evaluate the tiers of FORTIFIED homes as the mitigation strategies. I use Zillow ZTRAX and Institute of Business and Home Safety data to estimate how each level of FORTIFIED home designation affects property value. The results show FORTIFIED Gold designation on a new home has a 0%-8.4% increase on property value. I place my finding into a BCA of FORTIFIED designation to evaluate how this one benefit fits into the greater picture.