[pt] O cálculo do fluxo de potência em uma rede elétrica consiste em determinar o estado da rede, os fluxos e perdas elétricas nas linhas e as perdas internas totais no sistema. Nesse tipo de problema, a modelagem do sistema é estática e a rede é representada por um conjunto de equações e inequações algébricas. Diferentes métodos de solução foram propostos na literatura para realizar cálculos de fluxo de potência. No entanto, para redes de distribuição, esses métodos devem ser capazes de modelar, com detalhes suficientes, algumas características únicas desses sistemas, como sua estrutura quase radial, a natureza desequilibrada das cargas e a inserção de geradores distribuídos. Além disso, a modelagem do padrão de consumo nos sistemas de distribuição é mais complexa e os parâmetros das linhas são mais difíceis de serem obtidos, quando comparados com o sistema de transmissão. Portanto, a aplicação de métodos tradicionais para cálculos de fluxo de potência em redes de distribuição pode levar a soluções divergentes. Nesse contexto, o presente trabalho propõe uma nova abordagem para cálculos de fluxo de potência em sistemas de distribuição, baseada em Machine Learning. Os modelos propostos utilizam Redes Neurais Artificiais (RNAs) para prever as perdas ativas internas de uma rede de distribuição e os fluxos de potência nas fronteiras com o sistema de transmissão. Simulações numéricas demonstram o desempenho eficiente da abordagem proposta, além de suas vantagens computacionais em relação aos softwares normalmente utilizados nesse tipo de estudo pois, uma vez treinadas, as RNAs podem aproximar, de modo extremamente rápido, cálculos de fluxo de potência, já que apenas operações matriciais são realizadas. Além disso, o trabalho apresenta uma aplicação da metodologia proposta: as previsões, obtidas pela RNA, para os fluxos nas fronteiras com a rede de transmissão foram utilizadas para gerar contratos ótimos de demanda para um sistema de distribuição real no Brasil. / [en] The power flow calculation on an electric network consists of determining the network s state, power flows and electrical losses on the lines, and total losses on the feeder. In this type of problem, the system s modeling is static, and the network is represented by a set of algebraic equations and inequations. Different solution methods were proposed in the literature to perform power flow calculations. However, for distribution networks, these methods must be able to model, with sufficient details, some unique features of these systems, such as their near radial structure, the unbalanced nature of the loads, and distributed generators insertion. Besides that, modeling the consumption pattern in distribution systems is more complex, and the line parameters are more difficult to be obtained when compared to the transmission system. Hence, applying traditional methods for power flow calculations in distribution networks may lead to divergent solutions. Within this context, this work proposes a new approach for power flow calculations in distribution systems based on Machine Learning. The proposed models use Artificial Neural Networks (ANNs) to predict the active internal losses of a distribution network and the power flows at the borders with the transmission system. Numerical simulations demonstrate the effective performance of the proposed approach, as well as its computational advantages over benchmark software programs since, once trained, ANNs can approximate power flow calculations extremely fast, as only matrix operations are needed. Moreover, the work presents an application of the ANN methodology proposed: predictions of the flows at the borders with the transmission network were used to generate optimal demand contracts for a real distribution system in Brazil.

No description available.

This is probably the most appropriate time for the development of robust and reliable structural damage detection systems as aging civil engineering structures, such as bridges, are being used past their life expectancy and beyond their original design loads. Often, when a significant damage to the structure is discovered, the deterioration has already progressed far and required repair is substantial. This is both expensive and has negative impact on the environment and traffic during replacement. For the exposed reasons the demand for efficient Structural Health Monitoring techniques is currently extremely high. This licentiate thesis presents a two-stage model-free damage detection approach based on Machine Learning. The method is applied to data gathered in a numerical experiment using a three-dimensional finite element model of a railway bridge. The initial step in this study consists in collecting the structural dynamic response that is simulated during the passage of a train, considering the bridge in both healthy and damaged conditions. The first stage of the proposed algorithm consists in the design and unsupervised training of Artificial Neural Networks that, provided with input composed of measured accelerations in previous instants, are capable of predicting future output acceleration. In the second stage the prediction errors are used to fit a Gaussian Process that enables to perform a statistical analysis of the distribution of errors. Subsequently, the concept of Damage Index is introduced and the probabilities associated with false diagnosis are studied. Following the former steps Receiver Operating Characteristic curves are generated and the threshold of the detection system can be adjusted according to the trade-off between errors. Lastly, using the Bayes’ Theorem, a simplified method for the calculation of the expected cost of the strategy is proposed and exemplified. / <p>QC 20170420</p>

Alzheimer’s disease is a neurodegenerative disease that affects approximately 6% of the global population aged over 65 and is forecasted to become even more prevalent in the future. Accurately diagnosing the disease in an early stage can play a large role in improving the quality of life for the patient. One key development for performing this diagnosis is applying machine learning to perform computer-aided diagnosis. Current research in the field has been focused on removing assumptions about the used data sets, but in doing so they have often discarded objective metadata such as the patient’s age, sex or priormedical history. This study aimed to investigate the effect of including such metadata as additional input features to neural networks used for diagnosing Alzheimer’s disease through binary classification of magnetic resonance imaging scans. Two similar neural networks were developed and compared, one with these additional features and the other without them. Including the metadata led to significant improvements in the network’s classification accuracy, and should therefore be considered in future computer-aided diagnostic systems for Alzheimer’s disease. / Alzheimers sjukdom är en form av demens som påverkar ungefär 6% av den globala befolkningen som är äldre än 65 och förutspås bli ännu vanligare i framtiden. Tidig diagnos av sjukdomen är viktigt för att säkerställa högre livskvalitet för patienten. En viktig utveckling inom fältet är datorstödd diagnos av sjukdomen med hjälp av maskininlärning. Dagens forskning fokuserar på att ta bort subjektiva antaganden om datamängden som används, men har ofta även förkastat objektiv metadata såsom patientens ålder, kön eller tidigare medicinska historia. Denna studier ämnade därför undersöka om inkluderandet av denna metadata ledde till bättre prestanda hos neuronnät som används för datorstödd diagnos av Alzheimers genom binär klassificering av bilder tagna med magnetisk resonanstomografi. Två snarlika neuronnät utvecklades och jämfördes, med skillnaden att den ena även tog metadata om patienten som indata. Inkluderandet av metadatan ledde till en markant ökning i neuronnätets prestanda, och bör därför övervägas i framtida system för datorstödd diagnos av Alzheimers sjukdom.

The military recently recognized the importance of taking sociocultural factors into consideration. Therefore, Human Social Culture Behavior (HSCB) modeling has been getting much attention in current and future operational requirements to successfully understand the effects of social and cultural factors on human behavior. There are different kinds of modeling approaches to the data that are being used in this field and so far none of them has been widely accepted. HSCB modeling needs the capability to represent complex, ill-defined, and imprecise concepts, and soft computing modeling can deal with these concepts. There is currently no study on the use of any computational methodology for representing the relationship between adverse events and infrastructure development investments in an active war theater. This study investigates the relationship between adverse events and infrastructure development projects in an active war theater using soft computing techniques including fuzzy inference systems (FIS), artificial neural networks (ANNs), and adaptive neuro-fuzzy inference systems (ANFIS) that directly benefits from their accuracy in prediction applications. Fourteen developmental and economic improvement project types were selected based on allocated budget values and a number of projects at different time periods, urban and rural population density, and total adverse event numbers at previous month selected as independent variables. A total of four outputs reflecting the adverse events in terms of the number of people killed, wounded, hijacked, and total number of adverse events has been estimated. For each model, the data was grouped for training and testing as follows: years between 2004 and 2009 (for training purpose) and year 2010 (for testing). Ninety-six different models were developed and investigated for Afghanistan iv and the country was divided into seven regions for analysis purposes. Performance of each model was investigated and compared to all other models with the calculated mean absolute error (MAE) values and the prediction accuracy within ±1 error range (difference between actual and predicted value). Furthermore, sensitivity analysis was performed to determine the effects of input values on dependent variables and to rank the top ten input parameters in order of importance. According to the the results obtained, it was concluded that the ANNs, FIS, and ANFIS are useful modeling techniques for predicting the number of adverse events based on historical development or economic projects’ data. When the model accuracy was calculated based on the MAE for each of the models, the ANN had better predictive accuracy than FIS and ANFIS models in general as demonstrated by experimental results. The percentages of prediction accuracy with values found within ±1 error range around 90%. The sensitivity analysis results show that the importance of economic development projects varies based on the regions, population density, and occurrence of adverse events in Afghanistan. For the purpose of allocating resources and development of regions, the results can be summarized by examining the relationship between adverse events and infrastructure development in an active war theater; emphasis was on predicting the occurrence of events and assessing the potential impact of regional infrastructure development efforts on reducing number of such events.

Machine learning has been gaining popularity over the past few decades as computers have become more advanced. On a fundamental level, machine learning consists of the use of computerized statistical methods to analyze data and discover trends that may not have been obvious or otherwise observable previously. These trends can then be used to make predictions on new data and explore entirely new design spaces. Methods vary from simple linear regression to highly complex neural networks, but the end goal is similar. The application of these methods to material property prediction and new material discovery has been of high interest as many researchers have begun using the structure-property relationships of materials in conjunction with computational modeling to discover new materials with novel chemical and physical properties. One such class of materials is that of emulsion polymers, which are heavily used in the coatings industry as they serve as the binder in many waterborne coating systems. The great advantage of these materials is that they are synthesized in water at high solids (30-70%) and therefore are largely compliant with stringent environmental regulations. The chemistry of these polymers is highly variant, but the predominant chemistries include copolymers of styrene and acrylic monomers such as n-butyl acrylate or copolymers of only acrylic monomers. Due to the high degree of complexity and variability of these systems, modeling their behavior according to structure-property relationships is currently impractical. Instead, this thesis will demonstrate the use of supervised machine learning methods in conjunction with genetic algorithms to predict and optimize emulsion polymer performance based on recipe composition. These emulsion polymers will also be evaluated for use in concrete coatings meant to be applied with minimal preparation work, i.e. no etching.

Machine learning and neural networks have recently become very popular in a large category of domains, partly thanks to their ability to solve complex problems by finding patterns in data, but also due to an increase in computing power and data availability. Successful applications of machine learning include for example image classification, natural language processing, and product recommendation. Despite the potential upside of machine learning applied to financial data there exists relatively few articles published while the ones that do exist exhibit that there exists a potential for the tools that it provides. This thesis utilizes neural networks to price collateralized loan obligations which is a type of bond that is backed by a large pool of corporate loans, rather than being issued by a single company or government like a regular bond. The large pool of corporate loans and structure of a collateralized loan obligation makes it a good candidate for this type of research as it involves regressing a large number of variables into a final single real-valued price of the bond where the relations are not necessarily linear. The thesis establishes a relatively simple model and builds upon this using a state-of-the-art ensemble method while also exploring a volatility scaled loss function. The findings of this thesis are that artificial neural networks can price collateralized loan obligations using only their structural and loan pool data with an accuracy close to that of a human. Ensemble methods outperform non-ensemble methods and boost performance by up to 28% when only considering mean squared error while scaling the loss function with the inverse of market volatility does not boost performance. The best performing model can price a collateralized loan obligation tranche rated AAA with an average absolute error of 0.88 and an equity tranche with an average mean absolute error of 4.67. / Under de senaste åren har maskininlärning samt artificiella neurala nätverk blivit väldigt populära i många olika domäner. Detta är delvis tack vare deras förmåga att lösa komplexa problem genom att hitta mönster i data, men även tack vare en ökning i beräkningskraft samt att tillgängligheten av data har blivit bättre. Några exempel på områden där maskininlärning har applicerats med framgång är klassificering av bilder, språkteknologi samt produktrekommendationer. Trots att maskininlärning skulle kunna erbjuda en stor potentiell uppsida vid lyckad tillämpning på finansiella data finns relativt lite studier publicerade kring ämnet. De studier som däremot är publicerade visar på stora möjligheter inom området. Den här studien använder artificiella neurala nätverk för att prissätta ”collateralized loan obligations” (CLOs), som tyvärr inte har någon bra svensk översättning. En CLO utfärdar obligationer vars underliggande värde härstammar från en portfölj av företagslån, och är därmed ett finansiellt instrument. Strukturen av en CLO och dess underliggande lånportfölj ger upphov till en stor mängd data, vilket gör instrumentet till en bra kandidat för maskininlärning. Studien etablerar ett relativt enkelt neuralt nätverk som sedan används för ett jämföra med en ensemblemetod samt en modifierad loss funktion som tar höjd för volatilitet. Slutsatserna av den här studien är att neurala nätverk lyckas prissätta instrumenten näst intill lika bra som vad en människa skulle kunna göra med befintliga metoder som bygger på Monte Carlo simulering. Däremot är studiens metod inte lika beroende av antaganden som gör den befintliga metoden väldigt känslig. Vidare så bidrar ensemblemetoden som används till att minska det genomsnittliga felet i kvadrat med upp till 28%. Att ta höjd för volatilitet vid inlärning bidar inte till att minska felet.

Les récentes avancées dans le domaine de l'intelligence artificielle ont ouvert la voie au développement de nouveaux modèles d'activité cérébrale. Les réseaux neuronaux artificiels (RNA) formés à des tâches complexes, telles que la reconnaissance d'images, peuvent être utilisés pour prédire la dynamique cérébrale en réponse à une série de stimuli avec une précision sans précédent, un processus appelé encodage cérébral. Les jeux vidéo ont fait l'objet d'études approfondies dans le domaine de l'intelligence artificielle, mais n'ont pratiquement pas été utilisés pour l'encodage cérébral. Les jeux vidéo offrent un cadre prometteur pour comprendre l'activité cérébrale dans un environnement riche, engageant et actif, contrairement aux tâches essentiellement passives qui dominent actuellement le domaine, telles que la visualisation d'images. Un défi majeur soulevé par les jeux vidéo complexes est que le comportement individuel est très variable d'un sujet à l'autre, et nous avons émis l'hypothèse que les RNAs doivent prendre en compte le comportement spécifique du sujet afin de capturer correctement les dynamiques cérébrales. Dans cette étude, nous avons cherché à utiliser des RNAs pour modéliser l'imagerie par résonance magnétique fonctionnelle (IRMf) et les données comportementales des participants, que nous avons collectées pendant que les sujets jouaient au jeu vidéo Shinobi III. En utilisant l'apprentissage par imitation, nous avons entraîné un RNA à jouer au jeu vidéo en reproduisant fidèlement le style de jeu unique de chaque participant. Nous avons constaté que les couches cachées de notre modèle d'apprentissage par imitation parvenaient à encoder des représentations neuronales pertinentes pour la tâche et à prédire la dynamique cérébrale individuelle avec une plus grande précision que divers modèles de contrôle, y compris des modèles entraînés sur les actions d'autres sujets. Les corrélations les plus fortes entre les activations des couches cachées et les signaux cérébraux ont été observées dans des zones cérébrales biologiquement plausibles, à savoir les réseaux somatosensoriels, attentionnels et visuels. Nos résultats soulignent le potentiel de la combinaison de l'apprentissage par imitation, de l'imagerie cérébrale et des jeux vidéo pour découvrir des relations spécifiques entre le cerveau et le comportement. / Recent advances in the field of Artificial Intelligence have paved the way for the development of novel models of brain activity. Artificial Neural networks (ANN) trained on complex tasks, such as image recognition and language processing, can be used to predict brain dynamics in response to wide range of stimuli with unprecedented accuracy, a process called brain encoding. Videogames have been extensively studied in the AI field, but have hardly been used yet for brain encoding. Videogames provide a promising framework to understand brain activity in rich, engaging and active environments, in contrast to mostly passive tasks currently dominating the field, such as image viewing. A major challenge raised by complex videogames is that individual behavior is highly variable across subjects, and we hypothesized that ANNs need to account for subject-specific behavior in order to properly capture brain dynamics. In this study, we aimed to use ANNs to model functional magnetic resonance imaging (fMRI) and behavioral gameplay data, which we collected while subjects played the Shinobi III videogame. Using imitation learning, we trained an ANN to play the game closely replicating the unique gameplay style of individual participants. We found that hidden layers of our imitation learning model successfully encode task-relevant neural representations and predict individual brain dynamics with higher accuracy than various control models, including models trained on other subjects' actions. The highest correlations between layer activations and brain signals were observed in biologically plausible brain areas, i.e. somatosensory, attentional and visual networks. Our results highlight the potential of combining imitation learning, brain imaging, and videogames to uncover subject-specific relationships between brain and behavior.

With the ever-increasing adoption of renewable energy sources, the seamless integration of PV systems into existing grids becomes imperative. Therefore, this study investigates the integration of a PV-ESS system into sustainable urban living. It entails the development and evaluation of forecasting models for PV production and electricity consumption using artificial neural network models, as well as the analysis of linear optimization algorithms. These investigations give insight into the benefits, challenges, and implications of implementing a PV-ESS system. The photovoltaic generation forecasting model demonstrates high accuracy in winter months while encountering complexity in dynamic summer conditions. The model for estimating power demand poses challenges due to a variety of factors, including human behaviour and data quality.Moreover, the study focuses on the formulation and assessment of linear optimization models with two aims: minimizing costs and optimizing self-consumption. The first continually reduces electricity costs while increasing self-consumption, whereas the second maximizes self-consumption, with limitations in winter battery use. Finally, forecast precision appears as a crucial factor for optimization models. Forecast errors have an impact on the system’s operation. Improving forecasting accuracy and adaptive control strategies are therefore critical. / Med den ständigt ökande användningen av förnybara energikällor blir sömlös integration av solcellssystem i befintliga elnät nödvändig. Därför undersöker denna studie integrationen av ett solcellsenergilagringssystem (PV-ESS) i hållbart stadsboende. Det innefattar utveckling och utvärdering av prognosmodeller för solcellsproduktion och elförbrukning med hjälp av artificiella neurala nätverksmodeller, samt analys av linjär optimeringsalgoritmer. Dessa undersökningar ger insikt om fördelarna, utmaningarna och konsekvenserna av att implementera ett PV-ESS-system. Modellen för prognostisering av solcellsgeneration visar hög noggrannhet under vintermånaderna men stöter på komplexitet under dynamiska sommarförhållanden. Modellen för att uppskatta elförbrukning står inför utmaningar på grund av olika faktorer, inklusive mänskligt beteende och datakvalitet. Dessutom fokuserar studien på formulering och utvärdering av linjära optimeringsmodeller med två mål: att minimera kostnader och optimera självkonsumtion. Den första minskar kontinuerligt elkostnader samtidigt som den ökar självkonsumtionen, medan den andra maximerar självkonsumtionen med begränsningar i vinterbatterianvändning. Slutligen framstår precision i prognoser som en avgörande faktor för optimeringsmodeller. Prognosfel påverkar systemets drift. Därför är förbättring av prognosnoggrannhet och adaptiva kontrollstrategier avgörande.

User retention is a critical aspect of any business or service. Churn is the continuous loss of active users. A low churn rate enables companies to focus more resources on providing better services in contrast to recruiting new users. Current published research on predicting user churn disregards time of day and time variability of events and actions by feature selection or data preprocessing. This thesis empirically investigates the practical benefits of including accurate temporal information for binary prediction of user churn by training a set of Machine Learning (ML) classifiers on differently prepared data. One data preparation approach was based on temporally sorted logs (log-level data set), and the other on stacked aggregations (aggregated data set) with additional engineered temporal features. The additional temporal features included information about relative time, time of day, and temporal variability. The inclusion of the temporal information was evaluated by training and evaluating the classifiers with the different features on a real-world dataset from a MedTech application. Artificial Neural Networks (ANNs), Random Forrests (RFs), Decision Trees (DTs) and naïve approaches were applied and benchmarked. The classifiers were compared with among others the Area Under the Receiver Operating Characteristics Curve (AUC), Positive Predictive Value (PPV) and True Positive Rate (TPR) (a.k.a. precision and recall). The PPV scores the classifiers by their accuracy among the positively labeled class, the TPR measures the recognized proportion of the positive class, and the AUC is a metric of general performance. The results demonstrate a statistically significant value of including time variation features overall and particularly that the classifiers performed better on the log-level data set. An ANN trained on temporally sorted logs performs best followed by a RF on the same data set. / Bevarande av användare är en kritisk aspekt för alla företag eller tjänsteleverantörer. Ett lågt användarbortfall gör det möjligt för företag att fokusera mer resurser på att tillhandahålla bättre tjänster istället för att rekrytera nya användare. Tidigare publicerad forskning om att förutsäga användarbortfall bortser från tid på dygnet och tidsvariationer för loggad användaraktivitet genom val av förbehandlingsmetoder eller variabelselektion. Den här avhandlingen undersöker empiriskt de praktiska fördelarna med att inkludera information om tidsvariabler innefattande tid på dygnet och tidsvariation för binär förutsägelse av användarbortfall genom att träna klassificerare på data förbehandlat på olika sätt. Två förbehandlingsmetoder används, en baserad på tidssorterade loggar (loggnivå) och den andra på packade aggregeringar (aggregerat) utökad med framtagna tidsvariabler. Inklusionen av tidsvariablerna utvärderades genom att träna och utvärdera en uppsättning MLklassificerare med de olika tidsvariablerna på en verklig datamängd från en digital medicinskteknisk produkt. ANNs, RFs, DTs och naiva tillvägagångssätt tillämpades och jämfördes på den aggregerade datamängden med och utan tidsvariationsvariablerna och på datamängden på loggnivå. Klassificerarna jämfördes med bland annat AUC, PPV och TPR. PPV betygsätter algoritmerna efter träffsäkerhet bland den positivt märkta klassen och TPR utvärderar hur stor del av den positiva klassen som identifierats medan AUC är ett mått av klassificerarnas allmänna prestanda. Resultaten visar ett betydande värde av att inkludera tidsvariationsvariablerna överlag och i synnerhet att klassificerarna presterade bättre på datauppsättningen på loggnivå. Ett ANN tränad på tidssorterade loggar presterar bäst följt av en RF på samma datamängd.