Publikationer från KTH
Not a member yet
57933 research outputs found
Sort by
Comparative analysis of techno-economic and techno-environmental approach to optimal sizing and dispatch of hybrid solar–battery systems
No full textThis study investigates the techno-economic and techno-environmental performance of photovoltaic (PV) solar systems coupled with battery energy storage systems (BESS) in a Swedish context. The research uses mixed-integer linear programming (MILP) to optimise the dispatch strategy, minimising both operational costs and CO2eq emissions. By analysing grid signals, including electricity price and carbon intensity, the study determines the optimal size and operation of PV-BESS. The base case in Sweden was further examined by comparing the system with Italy and Poland, and by testing it with different load demand profiles. Italy and Poland were chosen due to their higher variability in grid price and carbon footprint, respectively, as well as more favourable solar conditions. The industrial and industrial shift profiles were chosen to assess the impact of load profiles with less variability compared to the base case residential profile. The key findings reveal distinct differences between economic and environmental optimisation, impacting system performance and highlighting the need for a balanced approach. Local conditions, such as grid signal volatility and solar PV production, are shown to significantly influence optimal system configurations. In Sweden, the economic approach led to higher system utilisation due to greater price volatility, while the environmental approach prioritised lower emissions. Additionally, the trade-offs between economic and environmental optimisation can lead to cost/environmental footprint increases between 25% and up to several times higher (up to 300 %). The study also finds that reducing the levelised cost of energy (LCOE) or levelised carbon footprint (LCO2eq) from the investor perspective may not always translate into significant end-user benefits. This further highlights the importance of including various stakeholder perspectives in the analysis, especially in the context of decision support. Sensitivity analysis indicates that oversizing the PV system leads to a rapid increase in costs and emissions. The addition of BESS can justify this increase by scaling the Renewable Energy Self-Sufficiency (RESS) value. Furthermore, there are diminishing returns for oversizing the battery. This research is relevant for various stakeholders, including project developers, policymakers, and researchers involved in renewable energy integration. Future research could further refine optimisation strategies for PV-BESS systems by delving deeper into specific aspects such as grid signal analysis and diverse end-of-life (EoL) pathways.QC 20250212</p
Röntgenabsorptiometri med multipla energier i fotonräknande detektorer - Metoder för förbättrad materialseparation
No full textDual-energy x-ray absorptiometry (DEXA) is a method that utilizes the dependence of the x-ray attenuation coefficient on photon energy and material to separate two materials, typically bone and soft tissue. With photon-counting detectors, it is possible to divide the detected x-ray spectrum into two energy bins during a single acquisition, producing two independent images of the same object. These images can be processed to subtract one material's contrast, leaving an image that shows only the contrast of the second material, as if the scan were done solely on it. Varex Imaging has developed a new generation of photon-counting detectors capable of dividing the spectrum into up to six energy bins. This advancement opens the door to multi-energy x-ray absorptiometry (MEXA). However, how to effectively utilize the additional energy information provided by these detectors remains unclear. Therefore, this thesis aims to develop an approach to utilize the additional energy information and assess whether the new method can outperform the traditional DEXA technique. Additionally, it explores whether the method can be extended to separate more than two materials, which is not possible with DEXA. Using both simulations and experiments, three new approaches that make use of the additional energy bins were developed: a filtered spectrum method, a mean-value approach, and a three-material separation method. All three methods relied on weighted logarithmic subtraction to address the problem. For the three-material separation, a novel 3D weighted logarithmic subtraction approach was introduced. The thesis concludes that the filtered spectrum method is equivalent to conventional DEXA, meaning it does not enhance material separation. However, the mean-value approach shows strong potential for accurately determining material thicknesses within an object. Additionally, the 3D weighted logarithmic approach successfully isolated two out of three materials in a three-material object, demonstrating that separating more than two materials is indeed possible.Dual-energy x-ray absorptiometry (DEXA) är en metod som utnyttjar attenueringskoefficientens beroende av fotonernas energi och material för att separera två material, vanligtvis ben och mjukvävnad. Med fotonräknande detektorer är det möjligt att dela upp ett detekterat röntgenspektrum i två intervall (energi-bins), vilket resulterar i två oberoende bilder av samma objekt. Dessa bilder kan sedan bearbetas för att subtrahera kontrasten från ett material, vilket resulterar i en bild bestående av enbart kontrasten från det andra materialet. Varex Imaging har utvecklat en ny generation av fotonräknande detektorer som kan dela upp spektrumet i upp till sex energi-bins. Denna utveckling öppnar upp möjligheter för multi-energy x-ray absorptiometry (MEXA). Det är dock fortfarande oklart hur den extra informationen som dessa detektorer tillhandahåller bäst kan utnyttjas. Därför syftar detta examensarbete till att utveckla en metod för att använda den extra informationen och utvärdera om den nya metoden kan överträffa den traditionella DEXA-tekniken. Dessutom undersöks om metoden kan utökas för att separera fler än två material, vilket inte är möjligt med DEXA. Både simuleringar och experiment genomfördes för att utveckla tre nya metoder som drar nytta av de extra energi-binsen: en “filtrerat ”spektrum-metod, en medelvärdesmetod och en metod för separering av tre material. Alla tre metoderna använde sig av viktad logaritmisk subtraktion (weighted logarithmic subtraction) för att lösa problemet. För separering av tre material introducerades en ny metod kallad “3D weighted logarithmic subtraction”. Slutsatsen av examensarbetet är att “filtrerat ”spektrum-metoden är likvärdig med konventionell DEXA, vilket innebär att den inte förbättrar materialseparationen. Däremot presterade medelvärdesmetoden väl och visar stor potential för att noggrant bestämma materialtjocklekar i ett objekt. Dessutom lyckades ”3D weighted logarithmic subtraction” isolera två av tre material i ett objekt bestående av tre material, vilket visar att det faktiskt är möjligt att separera fler än två materia
Focusing on Sustainability : Inspiration for Technology Education that Engages High school Students
No full textMed de pågående klimatförändringarna ökar behovet av hållbara lösningar, där tekniken spelar en central roll. Från och med 1 juli 2025 får hållbar utveckling en mer framträdande roll i gymnasiets nya läroplan, elever i teknikämnet ska ges möjlighet att utveckla förmågor att skapa tekniska lösningar för ett hållbart samhälle. I detta examensarbete genomförs en systematisk litteraturöversikt för att undersöka forskning om elevaktiviteter som kan främja gymnasieelevers engagemang för hållbar utveckling inom teknikundervisning. Resultaten från litteraturöversikten kommer att utgöra en grund för en lärarhandledning inom ämnena teknik och hållbart samhällsbyggande, med fokus på hållbar utveckling, vilken återfinns i bilaga A. Sju artiklar inkluderades efter genomförandet av den systematiska litteraturöversikten, och tre huvudteman identifierades utifrån hur aktiviteterna genomfördes: praktiskt byggande, rollspel och undersökning av närmiljöer. Samarbete med eller utan externa aktörer utgjorde ett undertema. Artiklarna diskuterades i relation till ARCS-modellen för att förstå hur aktiviteter kan utformas för att främja engagemang i teknikundervisning. Resultaten visade att aktiviteter som kopplar teknik till autentiska sammanhang, såsom deltagande i forskningsprojekt, samarbete med yrkesverksamma eller simulering av verkliga arbetsprocesser, är effektiva för att stimulera engagemang. Praktiska aktiviteter och problemlösning som ger konkreta resultat stimulerar också elevers engagemang. En viktig aspekt är att undervisningen ska kännas meningsfull, vilket kan uppnås genom att koppla aktiviteter till närmiljöer och visa på behovet av tekniska lösningar för en hållbar framtid. Slutsatserna pekar på behovet av forskning som föreslår aktiviteter med fokus på hållbar utveckling inom teknikundervisning, som främjar gymnasieelevers engagemang.With the ongoing climate change, the need for sustainable solutions is increasing, where technology plays a central role. Starting from July 1, 2025, sustainable development will have a more prominent role in the new curriculum for upper secondary school, with students in the technology subject being given the opportunity to develop the skills to create technological solutions for a sustainable society. This thesis conducts a systematic literature review to investigate research on student activities that can promote engagement for sustainable development within technology education. The results from the literature review will serve as the basis for a teacher's guide in the subjects of technology and sustainable community building, with a focus on sustainable development, which can be found in Appendix A. Seven articles were included after the systematic literature review, and three main themes were identified based on how the activities were carried out: practical building, role-playing, and exploration of local environments, with or without collaboration with external stakeholders as a sub-theme. The articles were discussed in relation to the ARCS model to understand how activities can be designed to foster engagement in technology education. The results showed that activities that link technology to authentic contexts, such as participation in research projects, collaboration with professionals or simulating real-world work processes, are effective in stimulating engagement. Practical activities and problem-solving that yield tangible results also stimulate student engagement. An important aspect is that the education must feel meaningful, which can be achieved by connecting activities to local environments and demonstrating the need for technological solutions for a sustainable future. The conclusions highlight the need for research which proposes activities in technology education focused on sustainable development that foster engagement in high school students
Approximation properties relative to continuous scale space for hybrid discretisations of Gaussian derivative operators
No full text This paper presents an analysis of properties of two hybrid discretisation methods for Gaussian derivatives, based on convolutions with either the normalised sampled Gaussian kernel or the integrated Gaussian kernel followed by central differences. The motivation for studying these discretisation methods is that in situations when multiple spatial derivatives of different orders are needed at the same scale level, they can be computed significantly more efficiently, compared to more direct derivative approximations based on explicit convolutions with either sampled Gaussian derivative kernels or integrated Gaussian derivative kernels. We characterise the properties of these hybrid discretisation methods in terms of quantitative performance measures, concerning the amount of spatial smoothing that they imply, as well as the relative consistency of the scale estimates obtained from scale-invariant feature detectors with automatic scale selection, with an emphasis on the behaviour for very small values of the scale parameter, which may differ significantly from corresponding results obtained from the fully continuous scale-space theory, as well as between different types of discretisation methods. The presented results are intended as a guide, when designing as well as interpreting the experimental results of scale-space algorithms that operate at very fine scale levels.QC 20250129Covariant and invariant deep network
Outlier-Robust Distributionally Robust Optimization via Unbalanced Optimal Transport
No full textDistributionally Robust Optimization (DRO) accounts for uncertainty in data distributions by optimizing the model performance against the worst possible distribution within an ambiguity set. In this paper, we propose a DRO framework that relieson a new distance inspired by Unbalanced Optimal Transport (UOT). The proposed UOT distance employs a soft penalization term instead of hard constraints, enabling the construction of an ambiguity set that is more resilient to outliers. Under smoothness conditions, we establish strong duality of the proposed DRO problem. Moreover, we introduce a computationally efficient Lagrangian penalty formulation for which we show that strong duality also holds. Finally, we provide empirical results that demonstrate that our method offers improved robustness to outliers and is computationally less demanding for regression and classification tasksQC 20250114</p
Regret Lower Bounds for Learning Linear Quadratic Gaussian Systems
No full textIn this article, we establish regret lower bounds for adaptively controlling an unknown linear Gaussian system with quadratic costs. We combine ideas from experiment design, estimation theory, and a perturbation bound of certain information matrices to derive regret lower bounds exhibiting scaling on the order of magnitude root T in the time horizon T . Our bounds accurately capture the role of control-theoretic parameters and we are able to show that systems that are hard to control are also hard to learn to control; when instantiated to state feedback systems we recover the dimensional dependency of earlier work but with improved scaling with system-theoretic constants, such as system costs and Gramians. Furthermore, we extend our results to a class of partially observed systems and demonstrate that systems with poor observability structure also are hard to learn to control.QC 20250121</p
Performance Optimization of Warehouse Management Systems: Identification and Elimination of Bottlenecks in the Scanning and Data Processing Flow : A technical analysis of system improvements and streamlining of digital workflows in warehouse management systems.
No full textPrestandaproblem i lagerhanteringssystem kan drabbas av märkbara fördröjningar vid skanning och databehandling, vilket påverkar effektiviteten i arbetsflöden och användarupplevelsen. Denna studie undersöker flaskhalsarna i ett befintligt lagerhanteringssystem och föreslår optimeringsstrategier för att minska svarstider. Arbetet inleddes med en uppdelad mätning av systemkomponenter såsom skanningsprocess, API- integration och databasåtkomst. Genom mätning av total API-svarstid och nätverkslatens identifierades att de längsta svarstiderna uppstod vid skrivoperationer mot det externa affärssystemet Microsoft Business Central. Som optimeringsåtgärd implementerades asynkron köhantering, vilket möjliggjorde snabbare återkoppling till användaren. Resultatet visar att de upplevda svarstiderna reducerades upp till 90 % i kritiska delen av processen. Arbetet bygger på en kombination av teknisk analys och praktiska verktyg, och kan därmed fungera som vägledning för framtida optimeringsprojekt inom liknande miljöer.Performance issues in warehouse management systems can be affected by noticeable delays in scanning and data processing, which affects the efficiency of workflows and the user experience. This study investigates the bottlenecks in an existing warehouse management system and suggests optimization strategies to reduce response times. The work began with a divided measurement of system components such as scanning process, API integration and database access. By measuring total API response time and network latency, it was identified that the longest response times occurred during write operations to the external business system Microsoft Business Central. As an optimization measure, asynchronous queue management was implemented, which enabled faster feedback to the user. The result shows that the perceived response times were reduced by up to 90% in critical parts of the process. The work is based on a combination of technical analysis and practical tools, and can thus serve as guidance for future optimization projects in similar environments
Mer än bara nyckelord : Förbättring av juridisk informationssökning via inbäddningar, korskodare och stora språkmodeller
No full textLegal Information Retrieval (LIR) is a specialized field focused on accessing legal texts, including legislation, case law, and scholarly works. Traditional keyword-based search methods often exhibit limitations due to challenges such as synonymy, polysemy, and the complex nature of legal language, leading to reduced recall and precision rates. To address these issues, this thesis aims to optimize the retrieval component within Retrieval-Augmented Generation (RAG) pipelines, ensuring that Large Language Models (LLMs) receive the most pertinent documents to accurately respond to user queries. We explore various methodologies, including embedding techniques, cross-encoders, and LLMs, to enhance the retrieval process. Embedding techniques involve transforming text into dense vector representations, facilitating efficient similarity searches. Cross- encoders jointly encode query-document pairs to assess relevance more precisely, albeit at a higher computational cost. By integrating these approaches, we aim to improve the selection of relevant documents, thereby enhancing the overall performance of RAG systems in the legal domain.Juridisk informationssökning är ett specialiserat område som fokuserar på att ge tillgång till juridiska texter, inklusive lagstiftning, rättspraxis och akademiska verk. Traditionella sökmetoder baserade på nyckelord uppvisar ofta begrän- sningar på grund av utmaningar som synonymi, polysemi och det komplexa juridiska språket, vilket leder till minskad täckning och precision. För att hantera dessa problem syftar denna avhandling till att förbättra täckningen inom Retrieval-Augmented Generation (RAG)-ramverk, för att säkerställa att stora språkmodeller (LLM) får de mest relevanta dokumenten för att exakt besvara användarfrågor. Vi utforskar olika metoder, inklusive inbäddningstekniker, cross-encoders och LLM, för att förbättra sökprocessen. Inbäddningstekniker in- nebär att transformera text till täta vektorrepresentationer, vilket underlättar effektiv sökning. Cross-encoders gör en gemensam kodning av sökfråga och dokument för att bedöma relevans mer exakt, om än till en högre beräkn- ingskostnad. Genom att integrera dessa metoder strävar vi efter att förbät- tra urvalet av relevanta dokument, och därigenom förbättra den övergripande prestandan hos RAG-system inom det juridiska området
Parameter study of potential active backward-facing step in diffuser of a passenger vehicle
No full textThis thesis investigates a passive flow-control device integrated into the rear diffuser of apassenger vehicle, with the primary aim of reducing aerodynamic drag while taking into accountlift effects. The study is purely performed using computational fluid dynamics (CFD) simulationsto evaluate how variations in diffuser angle and ride height affect the aerodynamic forces and why.Together with this, an active solution using a backward-facing step (BFS) geometry is investigatedby doing a parameter study. CFD results indicate that a diffuser angle approximately five degrees lower thanthe baseline provides the most effective drag reduction, with ride height showing a proportionalbut less pronounced influence on both drag and lift. The inclusion of a BFS demonstrated furtherdrag reduction in selected configurations, particularly when optimized for step angle, though it alsotends to increase lift in most cases. Based on the quantitative observations from the simulationadditional insights based on qualitative observations are provided into flow reattachment and wakebehavior, providing physical explainability to the results. Overall, the study concludes that passivegeometric modifications in the diffuser region with an active solution, especially the use of a BFS,can meaningfully enhance aerodynamic efficiency in road vehicles. However, these benefits must beweighed against potential increases in lift and their implications for vehicle stability if activated
Direct or indirect ownership? : A Qualitative Study of Swedish Institutional Investors' Strategies and Motives
No full textDenna uppsats undersöker de strategiska överväganden som ligger bakom svenska institutionella investerares val mellan direkt och indirekt ägande av fastigheter. Genom en kvalitativ ansats baserad på semistrukturerade intervjuer med yrkesverksamma från stora pensionsfonder och försäkringsbolag syftar studien till att identifiera argumenten för direkt fastighetsägande bland långsiktiga institutionella investerare, trots dess operativa komplexitet och resurskrav. Resultaten visar att de främsta motiven för att välja direkt ägande är ökad kontroll över investeringsstrategi, integrering av hållbarhetsaspekter (ESG), styrning samt kostnadseffektivitet. Direktägande, som vanligtvis sker genom hel- eller delägda fastighetsbolag, möjliggör för investerare att anpassa fastighetsförvaltningen till övergripande institutionella mål och samtidigt undvika vanliga intressekonflikter som kan uppstå i fondstrukturer. Indirekt ägande används främst för diversifiering, internationell exponering eller tillgång till nischade segment när den interna kapaciteten är begränsad. Studien visar att preferenser bland institutionella investerare formas av en kombination av faktorer såsom portföljstorlek, intern organisation, kapitalflöden och strategiska prioriteringar. Även om direktägande ofta föredras av större institutioner, utgör indirekta strukturer fortsatt ett viktigt verktyg för exponering inom områden där direkt ägande är mindre praktiskt genomförbart. Studien bidrar till en ökad förståelse för kapitalallokeringsbeslut i institutionella fastighetsportföljer och belyser hur valet av ägandeform speglar bredare organisatoriska, finansiella och styrningsrelaterade överväganden.This thesis investigates the strategic considerations behind Swedish institutional investors’ choice between direct and indirect real estate ownership. Through a qualitative approach based on semi-structured interviews with professionals from major pension funds and insurance companies, The study aims to identify the arguments in favor of direct property ownership among long-term institutional investors, despite its operational complexity and resource intensity. The findings reveal that the key motivations for choosing direct ownership include increased control over investment strategy, ESG integration, governance, and cost efficiency. Direct ownership, typically executed via wholly or jointly owned real estate companies, enables investors to align real estate management with their broader institutional goals while avoiding common agency issues associated with fund structures. Indirect ownership, in contrast, is mainly used for diversification, international exposure, or access to niche segments when internal capacity is limited. The study concludes that institutional investors’ preferences are shaped by a combination of factors such as portfolio size, internal organization, capital flows, and strategic priorities. Although direct ownership is often preferred among larger institutions, indirect structures remain an important tool for achieving exposure where direct engagement is impractical. The research contributes to the understanding of capital allocation decisions in institutional real estate portfolios, highlighting how the choice of ownership form reflects broader organizational, financial, and governance considerations