16168 research outputs found
Sort by
Investigation of lightning and commutation overvoltages in overhead lines.
No full textThis master's thesis investigates overvoltage processes caused by lightning discharges and switching operations in 110 kV overhead power transmission lines. Different types of overvoltages, their impact on the power grid, and potential protection devices are reviewed. Surge arresters are analyzed – their characteristics, selection of installation locations, and impact on the insulation of power transmission line equipment. The object of the study is a high-voltage 110 kV overhead line and its equipment, subjected to lightning overvoltages. Modeling of the overhead line and transformer substation is performed in the Electromagnetic Transients Program (EMTP) environment. Using the EMTP software package, lightning strikes to the overhead line and substation are simulated, determining overvoltage amplitudes and duration under different conditions. The most optimal parameters and installation location of surge arresters in the substation are determined
Benzimidazolo, tioksantono ir piridino fragmentų turinčių organinių puslaidininkių, skirtų organiniams šviesos diodams, sintezė ir tyrimas.
No full textOrganic light-emitting diodes offer a number of advantages. They can be used to produce thin and flexible displays with a brighter and richer colour gamut compared to competing devices. Organic light-emitting diode technology is one of the leading technologies, particularly in the manufacturing of smartphones, TVs and lighting devices. Efficient organic light-emitting diodes employ a variety of organic semiconductors in their emissive layers. Proper selection of such materials, ensuring compatibility between the energy levels of the host and the emitter, enables the efficient electricity-to-light conversion. One of the many advantages of organic semiconductors is the ability to easily modify their molecular structure by changing various fragments that affect their semiconducting properties. In this work, several series of emitters and hosts were synthesized and studied, detailing their synthesis, characterization and performance in prototype devices with external quantum efficiency reaching up to 30%
Aggregated LIME for the global explainability of classification models.
No full textAs machine learning (ML) models increasingly influence high-stakes decisions in healthcare, finance, and policy, the need for transparent and interpretable artificial intelligence (AI) systems becomes critical. Local Interpretable Model-Agnostic Explanations (LIME) offer a widely adopted approach for local model interpretability, yet their global aggregation remains methodologically limited. This project investigates and enhances global explanation strategies by aggregating local LIME explanations, addressing known limitations such as noise sensitivity and inconsistent feature relevance. To improve global interpretability, two novel strategies were tested: weighting explanations by their fidelity (R²) and kernel density estimation (KDE)-based aggregation. Experimental results confirmed that incorporating R² significantly reduced distributional divergence (JSD), improved stability, and mitigated performance degradation under feature removal. Although KDE achieved the highest rank correlation (Spearman = 0.91), it lagged in other metrics, indicating trade-offs. The enhanced aggregation methods outperformed existing literature baselines across several quality metrics, demonstrating their potential to strengthen the reliability and clarity of global explanations in classification models
Modelling and research of steam tunnel control systems.
No full textThis work presents a literature review on the structures of automatic control systems, PID controller structures, controller tuning methods, and the implementation of PID controllers in industrial controllers. The steam tunnel technological line and its technological process – steam heating of emulsion – are analyzed, along with its control systems: the steam flow control system for the upper hood and the temperature maintenance control system for the lower hood. Step response experiments are carried out to identify the transfer functions of the objects in the studied control systems. Models of the investigated control systems are developed using the MATLAB/Simulink software package. PI and PID controllers are tuned with reviewed control systems using various tuning methods. The existing control system is being modified in the model by changing its control. The research results and conclusions are presented
Prognozavimo spartinimas transformerių architektūroms.
No full textThis Master’s thesis focuses on improving inference performance in two transformer-based machine learning architectures. The first one is the Temporal Fusion Transformer (TFT), a strong baseline machine learning model architecture for multivariate, multi-horizon time series prediction. The second one is an encoder-decoder-based Large Language Model (LLM) such as Google’s Gemini. These models are usually trained on large-scale datasets, deployed at cloud scale, and served to millions of users concurrently. Because of this, an inference (i.e. model serving speed) is critically important. For the TFT model, the goal is to identify potential improvements, apply them, and compare the results against the original implementation. Similarly, for the LLM, the work explores how assisted generation techniques can be applied to reduce inference latency, following the same cycle of identifying, implementing, and evaluating improvements. The document is structured into four chapters. The first chapter presents the theoretical foundations and related works. The second outlines the proposed improvements and environments used for experimentation. The third details the experimental setups and results. The final chapter summarises findings and concludes the study
Lithuanian housing prices modelling with the use of big data analytics tools.
No full textThe dynamics of housing prices in Lithuania are highly important from both economic and social perspectives. Constantly rising prices not only erode household purchasing power but also result in housing expenses taking up an increasingly larger share of household budgets. Such market conditions compel individuals to carefully evaluate their housing choices, as these decisions are often tied to their most significant financial commitments. Failure to assess all alternatives and their potential can lead to serious consequences, including financial instability or even negative equity. Therefore, accurate, data-driven housing valuation is becoming increasingly important. While this study includes an overview of housing price indices and the macroeconomic factors influencing their trends, the primary focus is placed on modelling the value of individual real estate properties. The research involved collecting data from the real estate portal aruodas.lt, conducting exploratory data analysis, performing feature engineering, searching for optimal structural parameters, and developing regression models with tuned hyperparameters. Model performance was evaluated using RMSE and MAPE metrics, and the best results across all datasets were achieved using ensemble tree-based algorithms. The best-performing models were then used to conduct feature importance analysis. Using apartment sales data, it was determined that the most important explanatory variables for apartment square meter price were the building’s age and the Euclidean distance to the city center. Towards the end of the study, apartment price modelling was performed using segments derived from the most important explanatory variables. This allowed for the evaluation of model accuracy across different market niches and the development of practical recommendations for individuals planning to purchase an apartment. Additionally, rental prices were also modelled, enabling the calculation of approximate price-to-rent ratios for listed properties. These results provided insights into how long an investment in a particular apartment segment would take to pay off solely through rental income, thereby offering buyers an additional evaluation perspective that integrates the investment aspect of homeownership
Investigation of ABB robot manipulator control possibilities using virtual reality glasses.
No full textThis thesis explores the potential of virtual reality (VR) goggles for the control of robotic manipulators. The study provides an overview of the feasibility and effectiveness of using VR technology as a tool to improve the control interface of robotic systems. Prototype system has been developed to control an ABB robotic manipulator using a VR glasses application. By exploring robot control using this application, the aim is to identify the advantages and limitations of using VR glasses for robotic manipulation tasks
Research of dynamic export limitation of photovoltaic power plants based on power line loading.
No full textThe increasing integration of photovoltaic (PV) systems into electricity distribution networks presents new challenges related to voltage stability and line loading, particularly in low-voltage grids with prosumer installations. This research investigates the application of dynamic export limitation based on line loading to enhance grid reliability and support safe PV penetration in Lithuania’s distribution network. A simulation model was developed using “MATLAB” / “Simulink” to analyze voltage and power dynamics under varying irradiance and grid conditions. The model incorporated Maximum Power Point Tracking (MPPT) algorithms – Perturb and Observe (P&O) and Incremental Conductance (“IncCond”) – and was applied with both two-level and three-level inverters. The performance of these algorithms was compared under irradiance fluctuation and voltage disturbance scenarios. The results showed that the Incremental Conductance algorithm, combined with a three-level inverter, achieved faster stabilization (0.06 s) and a higher power factor (cos(φ) ≈ 0.917) than the P&O algorithm (0.14 s; cos(φ) ≈ 0.809). Additionally, the “IncCond” method exhibited lower power oscillations (~0.1 kW) and reduced harmonic amplitude (~10⁻⁴). Simulated power ranged from 2.65 to 5.84 kW, closely matching real measurements of 2.48 to 5.71 kW, with an average deviation of ±0.21 kW. Dynamic export limitation based on line loading and voltage thresholds (e.g., Vac > 250 V) proved to be an effective strategy to reduce overvoltage risk and improve system flexibility. The results validate the model’s accuracy for real-world analysis and highlight the importance of integrating such methods into future grid management frameworks to support renewable energy expansion in Lithuania
Žiedinės ekonomikos priemonių diegimas Jonavos savivaldybėje taikant ,,Gyvosios laboratorijos" metodą.
No full textAs urban areas increasingly contribute to environmental challenges, it is essential to integrate sustainable development goals to improve their resilience and sustainability. For cities to experience the advantages of these initiatives, they need to operate as communities that collaborate to achieve environmental objectives, partnering with researchers, government entities, and the business sector. This research seeks to explore the feasibility of integrating circular economy practices into urban management systems in Jonava. To accomplish this objective, a literature review is conducted, a material flow analysis of the Jonava municipality's waste management system is modeled using the STAN program, budget allocations for sectors are assessed through Microsoft Excel, an LCA is performed with CCaLC2 program, and discussion and feedback is done with the public organisations. The focus of this research is on 39 public organisations and institutions located in the Jonava municipality, which have been categorised into seven sectors: education, health, social services, sports, municipal services, public governance, and culture. Data for this analysis is derived from public procurement records for the year 2023. This project evaluates the environmental impact of purchased goods, alongside electricity, heating, water, and waste management. Findings indicate that Jonava municipality recycles only 2 thousand tons of waste annually, with a total of 10 thousand tons classified as mixed municipal waste, highlighting opportunities for improved sorting methods. In terms of budget allocation across the examined sectors, it has been determined that more than half of the total budget is dedicated to the Education and Municipal Services sectors, and for most sectors, the largest share of the budget is spent on product purchases. The functional unit used for the life cycle assessment represents the items purchased by each sector for the year 2023. Each sector possesses a unique functional unit due to the varying quantities of items purchased. The results reveal that the most significant environmental impacts are found in the Municipal Services sector (primarily from energy and fuel), the Education sector (largely resulting from food products and electronics), and the Sports sector (mainly due to wastewater treatment), indicating major areas for potential improvement for the municipality. Other recognised challenges in adopting circularity initiatives include a lack of awareness, difficulties in altering behavior, and insufficient funding for circularity efforts. Suggested actions for the Jonava municipality include enhancing energy efficiency (particularly in the Education and Municipal Services sectors), modernizing transportation and optimizing driving routes (especially in the Municipal Services sector), establishing self-catering systems (for the Education and Social Services sectors), and promoting electronic document storage rather than printing (for the Education and Municipal Services sectors)
Enzymatic modification of apple pomace and its application in conjunction with probiotics for jelly candy production /
No full textThis study aimed to evaluate the applicational possibilities of enzymatically modified apple pomace (AP) in conjunction with probiotics as value-added ingredients for the production of jelly candies. AP was enzymatically modified with Pectinex® Ultra Tropical, Viscozyme® L, and Celluclast® 1.5 L (Novozyme A/S, Bagsværd, Denmark), and the soluble and insoluble dietary fibre content was determined using the Megazyme kit (Megazyme International Ireland Ltd., Wicklow, Ireland), reducing sugar content using the 3,5-dinitrosalicylic acid assay. The technological properties of the modified AP, such as its swelling capacity, water-retention capacity, oil-retention capacity, bulk density, and static and thermal emulsion stability, were evaluated. Enzymatically modified AP hydrolysed with Celluclast® 1.5 L was used for the production of jelly candies supplemented with Bifidobacterium animalis DSM 20105. The survival of probiotics in the jelly candies during in vitro digestion, the viability of probiotics during candy storage, and candy quality characteristics were analysed. Enzymatically modified AP had different carbohydrate compositions and technological properties, depending on the enzyme preparation used. Although the viability of probiotics in the jelly candies decreased during storage, a significantly higher viability of B. animalis was determined in jelly candies supplemented with hydrolysed AP compared with control candies made without AP after digestion in the saline, gastric, and intestine phases. This study shows that Celluclast® 1.5 L can be used for increasing the soluble dietary fibre in AP (18.4%), which can be further applied, in conjunction with B. animalis, for added-value jelly candy production