82837 research outputs found
Sort by
Study On Electricity Market Dynamics, Cycling And Emissions In Decarbonized Scenarios Of The Alberta Electricity Market
No full textThis research explored electricity decarbonization through two approaches: simulating scenarios with increased variable renewable energy (VRE) integration using historical market data and employing an energy market simulation model to identify generator fleet mixes that maximize system value.
Focusing on Alberta’s electricity market, the study analyzed photovoltaic (PV) energy’s impact on emissions, load and pricing during the province’s transition period from coal to natural gas. Simulations from 2010 to 2019 found that increasing levels of PV capacity (from 15 MW to 3000 MW in a system with an average load of around 9000 MW) significantly reduced greenhouse gas emissions. As the carbon price increased from 15 to 30 Canadian dollars per tonne of carbon dioxide equivalent (CAD/tCO₂e), it significantly impacted the merit order by prioritizing the displacement of higher-emitting sources like coal, enhancing the environmental benefits of PV energy. Consequently, the displacement of coal-fired generation emissions by PV energy rose from 6.1% in 2010 to 10.6% in 2019, while the displacement of natural gas (NG)-fueled generation decreased from 30.2% to 6.9% over the same period. However, inserting modelled PV output into existing market merit orders resulted in steep midday price declines. At low installed capacities, PV plants achieved high market values because their midday production aligned with peak electricity prices. However, as PV capacity increased, oversupply during high solar production hours drove electricity prices down, drastically reducing PV's market value. With 1 GW of installed PV capacity, market value decreased by 30%, 66%, and 78% under low, medium, and high price regimes, respectively. At 3 GW of capacity, the declines were even more pronounced, reaching 51%, 75%, and 95%, respectively. Furthermore, the research explored optimizing PV system orientations (panel’s tilt and azimuth angles) to maximize revenue while addressing aforementioned price cannibalization. Simulations of PV capacities found that, at low capacities, energy-maximizing and revenue-maximizing orientations aligned. At higher PV capacities, revenue-optimal orientations shifted to times of higher-value energy but lower total generation due to intensified price cannibalization. Incorporating carbon credits further aligned revenue and energy-maximizing configurations.
Lastly, the effects of carbon pricing on the cycling behavior of fleet mixes with substantial shares of net-zero resources in Alberta by 2035 were studied. The research simulated electricity market operations using commitment and dispatch modeling. Cycling events, such as ramps and startups, were influenced not only by the variability of wind and solar (W&S) energy but also by variable-cost competition between NG-fueled generation without CCS and net-zero resources. Emission costs, driven by the carbon price, introduce a premium that can alter the dispatch order of generators, prioritizing lower-emission resources and reshaping the operational dynamics of the electricity market. Carbon pricing incentivized NG-fueled generation with carbon capture and sequestration (CCS) over NG without CCS, reducing emissions and cycling events. W&S-dominated fleet mixes achieved emissions intensities of 67 kilograms of carbon dioxide equivalent per megawatt-hour (kgCO₂e/MWh), outperforming non-W&S NG-dominated mixes of ~100 kgCO₂e/MWh. Blue hydrogen competed in the merit order against NG without CCS at an estimated carbon price of 170 CAD/tCO₂e, while nuclear facilities would have dispatch priority over thermal units with hydrogen or CCS due to its low variable operating costs.
The electricity sector faces uncertainty about how fleet mixes will evolve to reduce emissions, raising concerns among society and stakeholders about costs, revenues, and operational impacts. This dissertation examined key variables, including the effects of PV energy on market prices, how market price dynamics impact PV revenues, and how PV facilities can adapt to changes in fleet composition. It also explored future fleet mixes with net-zero resources, analyzing scenarios with and without W&S and the operation of hydrogen, natural gas with CCS, and nuclear energy. Additionally, it evaluated the impact of carbon pricing on PV revenues and operation patterns of net-zero resources such as dispatch and cycling (ramps and starts). By presenting various net-zero integration scenarios, this research serves as a valuable reference for policymakers and society, aiding informed decision-making in the energy transition
Predicting Large Customer Load Behaviour and Impact on Distribution Feeders using Clustering and Classification for Distribution Planning
No full textThe modern electrical distribution grid is at a critical juncture, tasked with accommodating unprecedented and often unpredictable load growth from the widespread electrification of transport and heating. This thesis presents a data-driven framework to address a fundamental challenge in electricity utility distribution planning: characterizing and predicting the grid impact of new large commercial and industrial (C&I) customer connections. Traditional planning models, which have long relied on a deterministic toolkit of heuristics, such as applying a static coincidence factor to a customer's nameplate rated power, are proving increasingly inadequate. This reliance on generalized, top-down assumptions often leads to suboptimal capital expenditure, either through the costly oversizing of assets to handle a worst-case scenario that rarely materializes, or, more dangerously, the underestimation of network risk, which can compromise grid reliability. The emergence of Advanced Metering Infrastructure (AMI) provides a path forward, offering a wealth of high-resolution consumption data that enables a more granular, behavior-based approach to load characterization.
This thesis develops, validates, and demonstrates a complete end-to-end process pipeline that translates this raw smart meter data into actionable, risk-aware planning insights. The methodology utilizes a robust two-stage K-Means clustering approach applied to a comprehensive, six-year proprietary dataset of seasonal, 15-minute interval consumption data from over 1,700 C&I customers. In the first stage, a large quantity of individual daily load profiles are clustered to identify a set of fundamental daily shape profiles. In the second stage, each customer is profiled for a given season by a feature vector representing their frequency of exhibiting each profile shape. This customer-level vector is then clustered to produce the final seasonal archetypes. While cutting-edge deep learning methods are available, the K-Means algorithm was deliberately selected for its optimal balance of scalability, proven robustness, and, most critically, the high interpretability of its resulting centroids. This choice was empirically validated in this study, where standard K-Means was found to outperform a more complex Variational Autoencoder-based approach on this dataset according to multiple clustering validity indices.
A key empirical contribution of this thesis is the longitudinal analysis of these customer segments. It was found that the proportional size of the identified clusters remained remarkably consistent over the six-year period and across summer and winter seasons, providing strong evidence of consistent, underlying behavioral patterns. Furthermore, this thesis demonstrates that cluster membership can be accurately predicted (with 82–87% accuracy for the two largest segments) using a classification framework built upon a minimal set of readily available static metadata. This is a significant finding for utility planners, as it avoids the need for costly surveys or external data acquisition. Permutation importance analysis identified rated power (kW) as the dominant predictor. The temporal analysis also yielded novel insights, identifying two critical customer subgroups: a persistent group of 'under-utilizing' customers whose low actual consumption belies their high rated power, and a smaller but important cohort of 'dynamic' customers who migrate between behavioral archetypes year-over-year. These groups represent valuable targets for demand-side management programs, tariff review, and data quality investigations.
The primary contribution of this work is the integration of these analytical stages into a practical planning application that bridges the gap between data science and engineering practice. A Monte Carlo simulation framework was developed, which uses stochastic profiles generated from the validated clusters to perform a probabilistic grid impact analysis. This transforms the planning input from a single, deterministic peak load value into a full probability distribution of potential future feeder loading. The resulting probabilistic output allows planners to move beyond a binary 'pass/fail' assessment to a nuanced, risk-quantified decision-making process. Ultimately, this research provides utility planners with a sophisticated, defensible, and repeatable analysis method to more accurately justify prudent capital investment, assess overload risk with a specified confidence level, and navigate the complexities of modern grid planning in a real-world business environment
Intraoral Diagnostic Ultrasound to Assess Gingival Thickness Measurement
No full textAbstract
Background: Assessment of periodontal parameters such as gingival thickness is relevant during diagnosis and treatment planning in all disciplines of dentistry. Thin and thick gingival thickness respond differently to inflammation, restorative trauma, and surgical insult. Current methods of evaluating gingival thickness include transgingival probing (TGP), visual assessment (VA), and probe visibility (PV) methods. Ultrasound (US) has been shown to be an accurate tool to assess periodontium, including gingival thickness. Evidence comparing the diagnostic performance of ultrasound against conventional methods for accurate classification of periodontal gingival thickness remains limited. Objectives: The primary objective of our study was to assess the performance of US gingival thickness measurements compared to transgingival probing as the gold standard. The secondary objective was to assess US gingival thickness measurements compared to probe visibility and visual assessment methods. Materials and Methods: This prospective diagnostic accuracy study included adult patients from the Graduate Periodontics clinic at the Oral Health Clinic, Mike Petryk School of Dentistry, University of Alberta. Maxillary central incisors of these patients were considered for the study. The US imaging was performed using a 20MHz intraoral transducer, and the gingival thickness was measured from three high-quality B-mode images, and the average was recorded. TGP was performed using a #8 endodontic file with a stopper under topical anesthesia, with thickness measured by digital callipers. VA was performed with the patient sitting in the upright position. This was followed by PV assessment using Colorvue gingival probe) (HuFriedy, Chicago, Illinois). Triplicate measurements were obtained for TGP and US. Diagnostic accuracy, agreement, and reproducibility were evaluated using Intraclass Correlation Coefficient (ICC), Bland-Altman analysis, non-inferiority testing, and Fisher’s Exact Test. To assess the performance of VA and PV, we compared them to the gingival thickness category of their numerical counterparts (TGP and US). Sensitivity and specificity were calculated using a 1.46 mm threshold to define thin gingiva.
Results: Of the 34 participants recruited, 31 completed all assessments. TGP and US classified the majority of cases as having thin gingiva (<1.46mm). US demonstrated excellent agreement with TGP when triplicate measurements were averaged (ICC(3,k) =0.91). Non-inferiority testing confirmed that US was not inferior to TGP (p = 0.0448), and Bland-Altman analysis showed a slight underestimation by US (mean difference = -0.0663mm, p = 0.246, Student’s t-test). Fisher’s Exact test revealed no significant association between VA, PV-white, PV-green or PV-blue (p = 1.00), while PV-none showed a significant association (p = 0.0065). VA and PV methods showed poor sensitivity in identifying thin gingiva (10.3-10.7% and 7.1%-89.3% respectively). US demonstrated the highest sensitivity for detecting thin gingiva (96.4%). VA and PV-white showed 100% specificity, while PV-green showed 33.3-66.7%, PV-blue showed 0% and US showed 66.6% specificity at detecting thick gingiva. Conclusion: Within the limits of our study, we can say that the US method of measuring gingival thickness is comparable to the gold standard (TGP). US performs superiorly to VA and PV, making it a potential non-ionizing, real-time tool to assess gingival thickness in patients
Leveraging Implementation Science in Learning Health Systems
No full textMany hail using implementation science (IS) as a key activity of high functioning learning health systems (LHSs). Even though IS is considered a LHS accelerant, little guidance exists to support embedding IS studies and applying knowledge generated from IS into everyday LHS functions. In the last decade, Canadian LHSs have emerged under various models. Yet the field remains nascent, with literature often focusing on the implementation and health outcomes of single projects or initiatives. The objective of this dissertation was to generate evidence of how to leverage IS and apply the derived results to support various LHS goals. This dissertation includes four studies. I conducted a social network analysis of IS partnerships in a province-wide learning health system in Alberta, Canada. Qualitative interviews accompanied this study to contextualize the social network analysis results and provide recommendations to strengthen IS collaborations opportunities in LHS settings. I also explored two models for conducting and applying implementation research in LHSs. A comparative case study highlighted how to use the community of practice model to support implementation research in LHS. A longitudinal case study of the Alberta IS Collaborative highlighted how to co-design infrastructure to support the conduct and use of IS in an established LHS. I designed the four studies to generate evidence to support the overarching dissertation objective. The analyses generated important practical insights into how to build foundational implementation research-practice partnerships and infrastructures to leverage IS in LHS settings. Specifically, the studies in this thesis provide evidence for assessing LHS capacity to embed and leverage implementation research-practice partnerships to strengthen LHS activities
Barriers to Needle Exchange Success in Canadian Prisons
No full textIn 2018, Correctional Service Canada rolled out the Prison Needle Exchange Program (PNEP) in response to growing human rights and public health concerns surrounding the spread of HIV and HCV in Canadian prisons. Despite having goals to save lives and reduce harm, an independent program evaluation in 2020 revealed that numerous prisons had zero program participants and, as of 2022, fewer than fifty individuals were participating in the program. In 2019, members of the University of Alberta Prison Project (UAPP: PI Drs. Bucerius and Haggerty) interviewed 15 prison staff and 63 incarcerated women and surveyed 74 prison staff at one of Canada’s federal women’s prisons regarding their views of substance use and harm reduction strategies. Using a mixed-methods research approach, I conduct a thematic analysis of these qualitative interviews to identify perceptions toward the prison needle exchange program. Further, I augment qualitative insights with findings from the institution-wide survey of 74 prison staff, including 38 correctional officers and 36 other staff.
In Paper 1, I use an implementation science research perspective, which aims to translate research findings into practice, identifying numerous barriers to the PNEPs’ successful implementation. My findings reveal an overwhelming lack of support toward the PNEP for both prison staff and incarcerated women. Interviews reveal three main themes for such disapproval: organizational barriers to program uptake, subcultural barriers to program uptake, and health and safety concerns.
In Paper 2, I run a stepwise linear regression analysis of 74 prison staff to understand what factors influence staff approval toward the PNEP, with several notable results. My findings show significant differences between correctional officers and other staff, with correctional officers being more disapproving toward the PNEP. Interestingly, prison staff’s familiarity with harm reduction programs has a negative relationship with PNEP approval, with further analysis revealing this relationship is only significant for correctional officers. This finding suggests a different understanding of “familiarity” between correctional officers and other staff, which may be informed by their unique lived experiences and work responsibilities. Perceptions of harm reduction program effectiveness has a significant positive relationship with PNEP approval, which suggests the more effective staff perceive programs, the higher approval they will have toward such programs.
Overall, my research reveals significant barriers to the PNEPs’ success, identifying a lack of support from both incarcerated women and correctional staff. These findings are consistent with previous literature, which highlight numerous problems with PNEP design and implementation. I conclude with recommendations for future research and possible reforms to improve implementation fidelity moving forward
Introduction to Community Support Work
No full textThis open educational resource (OER) came about through collaboration from the Community Support Worker team at NorQuest College. As a team, we wanted to develop an all encompassing resource for students in the Community Support Worker program that includes theory and elements of practice.
Four specializations make up the core of this program: Disability, older adults, Indigenous Peoples, and newcomers. As the Community Support Worker program explores working within diverse communities, particularly communities that have experienced marginalization, we need to acknowledge the impact of colonization, systemic barriers, trauma, poverty, discrimination, addiction, and involvement in the criminal justice system that many of these communities and community members face.
This book presents the helping skills required to work effectively and meaningfully with a diverse range of clients. Focusing on knowledge of social issues, compassion and empathy, interpersonal communication skills, self-awareness, and critical thinking helps to develop an understanding of tolerance and advocacy as a helper in the human services field. We hope this book provides a foundation for students new to the field as well as enhancing the approach to helping for those already working in the field. At the end of each chapter you will find both a scenario and a case scenario to test your learning and challenge your perspective. At the end of the textbook, you will find additional resources, including one scenario, three case scenarios, and fifteen practice questions for each chapter.
This book discusses key helping skills and showcases how to put those skills into practice with clients to provide solutions and knowledge of available resources. These skills are explained through theory, and we highlight the key social issues and barriers faced by clients. We include a section on trauma, intergenerational trauma, and becoming a trauma informed practitioner, which is important when working with clients. This book provides a free resource for students to become well-informed on the role and practice of a community support worker, to develop knowledge in areas of potential employment, and to instill the inner self-awareness, reflection, and critical thinking skills necessary to provide the best possible service.
Mi’Kmaq Elder Albert Marshall coined the phrase etuaptmumk “Two-Eyed Seeing” that is used worldwide. This refers to “learning to see from one eye with the strengths of Indigenous knowledges and ways of knowing, and from the other eye with the strengths of Western knowledges and ways of knowing… and learning to use both eyes together, for the benefit of all” (Institute for Integrative Science & Health, n.d.)
Damage Mechanisms in Steam Generation and Distribution Systems of Steam Assisted Gravity Drainage (SAGD)
No full textIn Oil Sands SAGD (Steam Assisted Gravity Drainage) operations, steam is injected into underground bitumen deposits to lower bitumen viscosity, allowing it to flow. Depending on the design of the facility, the steam distribution system (SDS) can include hundreds of meters or even kilometers of piping/pipeline transporting steam from the once through steam generators (OTSGs) to the well pads. The OTSGs and pipelines in SDS have been identified to undergo damage and failure after being in the service for few years. This thesis reports the failure investigations conducted on OTSG and SDS pipelines with the morphology, distribution and features of the damage identified. The associated process conditions, potential damage mechanisms identified, and their contributing factors are also reported. Once through steam generators boiler tubes analyzed experienced localized metal loss in the form of pitting corrosion, under deposit corrosion, micro-galvanic corrosion and, caustic and acidic attack from the operations. Some OTSG tubes have been found to not adhere to the required material composition expected for ASME SA-213 T11 standards: high Mn, high C, and low Cr. Sections of an HP-SDS pipeline have been ascertained to undergo pitting corrosion, crevice corrosion in the form of under-deposit corrosion, flow assisted corrosion and micro-galvanic corrosion. Results from the lab scale experiments conducted to provide an increased understanding on passivation of steels used in the SAGD process conditions are also presented. Recommendations in terms of (a) design and operating factors, (b) environment and operating factors and (c) manufacturing, fabrication and assembly factors to lower the number of failures and reduce the rates of degradation identified are provided
Automated process planning and decision-support framework for 3D food printing: a case of chocolate-based materials
No full textThe advancement of 3D food printing technology represents a significant innovation in the food manufacturing industry, offering unprecedented customization and efficiency. This
thesis presents an autonomous process-planning framework developed specifically for 3D food printing, addressing three critical research objectives that enhance and optimize the
production and efficiency of 3D food printing at large. I use chocolate-based materials as a reference material to illustrate the outcomes of each objective in this thesis. The main goal
of the thesis is to develop a framework for autonomous process planning to address the three research gaps detailed below.
The first objective focuses on enhancing printability through design features analysis by examining the intrinsic characteristics of the 3D designs and their influence on the
manufacturing process. This piece of research develops an optimization method to improve the structural integrity and aesthetic quality of printed food items such as chocolates, thus
improving the overall quality of the actual print. This involves a comprehensive analysis of design parameters and machine capabilities. Key aspects include the investigation of the
complexity of 3D designs, machines’ ability to manufacture a given design, and the classification of printability based on design features. The study produces a set of guidelines and tools that enable more reliability for better quality food printing outcomes, ensuring that printed structures maintain their intended shapes and textures post-printing.
The second objective is the design and development of a printability decision-support system. This system aids users in making informed decisions and optimizing the printing settings of complex food designs created through the extrusion-based techniques of 3D food printing. By incorporating an experimental result to build predictive models, the decision-support system provides optimized decisions based on real-time analysis to achieve the highest possible quality standards and aesthetic expectations of the final products. This system utilizes machine learning (ML) techniques to predict potential manufacturing issues and suggests adjustments to the printing settings to mitigate these problems. It is an intelligent knowledge-based system that guides users in achieving optimal results.
The third objective explores the development of job-scheduling models for autonomous 3D food printing, applicable to single and multiple-machine environments. Efficient job scheduling is crucial to maximizing throughput while minimizing processing time. This part of the research integrates advanced algorithms using the Mixed-Integer Linear Programming (MILP) approach to optimize the job scheduling process that can dynamically adjust to varying production demands and multiple constraints. This job-scheduling model ensures optimal utilization of resources, reduces downtime, and enhances overall production efficiency. It addresses challenges such as machine characteristics, part prioritization, and resource allocation to obtain flexible and scalable solutions that adapt to real-time changes in the 3D food printing environments.
Overall, this thesis contributes to the advancement of 3D food printing by addressing key challenges in design printability, material printability, and job-scheduling optimization. The
proposed autonomous process-planning framework improves the efficiency and reliability of 3D food printing and allows for broader adoption and innovation in the food industry. The
findings from this research have significant implications for the development of next-generation food printing systems, enhancing their capability to produce complex and high-
quality food products with greater precision and consistency
Contrastive Parallel Denoising for Improving Attribute Alignment of Diffusion models
No full textDiffusion models (DMs) have emerged as a powerful class of generative models, demonstrating remarkable performance in high-quality image synthesis, particularly in text-to-image generation. Leveraging iterative denoising steps guided by textual input, these models have enabled significant advancements in controllable image generation across a wide range of applications—from artistic content creation to data augmentation and virtual scene generation. However, despite their success, diffusion models continue to face major challenges in semantic alignment, especially when tasked with rendering complex prompts that involve multiple objects, attributes, or intricate spatial relationships.
A recurring issue in this setting is semantic misalignment, where generated images fail to accurately reflect the detailed semantics of the prompt. This often manifests in several ways: missing objects, undesired object fusion (e.g., two distinct objects blending into one), or incorrect object-attribute pairings (e.g., a red shirt rendered on the wrong character). Such errors are particularly common in prompts that contain several object-attribute bindings or hierarchical structures. One key contributing factor is the inter-contamination of textual embeddings during the diffusion process. Since most models encode the full prompt into a global representation, distinctions between different semantic components may become blurred, especially in the later denoising steps.
To address these limitations, this thesis introduces Contrastive Parallel Denoising (CPD)—a training-free framework designed to improve object-attribute alignment and mitigate semantic entanglement in diffusion-based text-to-image generation. CPD decomposes a complex prompt into multiple sub-prompts, each corresponding to a distinct object or concept. These sub-prompts are independently embedded and passed through parallel denoising branches. To promote semantic separation and reduce feature blending, we propose two novel contrastive mechanisms: the \textit{intra-object} contrastive loss, which encourages that different instances of the same object across branches remain consistent, and \textit{inter-object} contrastive loss, which ensures dissimilarity between distinct objects and attributes in the latent space. Both contrastive objectives act directly on the denoised latent representations at multiple stages, helping to preserve structural integrity and semantic clarity. Finally, the resulting latents are merged into a single coherent image using attention-based fusion strategies guided by cross-attention maps from the original full prompt.
We evaluate the proposed approach on CompBench, a benchmark specifically designed to test compositional understanding in generative models. CPD achieves substantial improvements over existing training-free diffusion techniques, as measured by BLIP-VQA scores and human preference ratings, showing more accurate object rendering, better attribute localization, and reduced feature entanglement.
This thesis contributes a practical and effective solution to a long-standing problem in generative modeling: aligning complex natural language prompts with faithful visual outputs. By offering a training-free alternative, CPD not only maintains the strengths of existing pre-trained diffusion models but also extends their utility to more demanding generation scenarios without additional computational overhead or retraining
