No description available.

Risk assessment is essential for nuclear power plants (NPPs) due to the complex dynamic nature of such systems-of-systems, as well as the devastating impacts of nuclear accidents on the environment, public health, and economy. Lessons learned from the Fukushima nuclear accident demonstrated the importance of enhancing current risk assessment methodologies and developing efficient early warning decision support tools. Static probabilistic risk assessment (PRA) techniques (e.g., event and fault tree analysis) have been extensively adopted in nuclear applications to ensure NPPs comply with safety regulations. However, numerous studies have highlighted the limitations of static PRA methods such as the lack of considering the dynamic hardware/software/operator interactions inside the NPP and the timing/sequence of events. In response, several dynamic probabilistic risk assessment (DPRA) methodologies have been developed and continuously evolved over the past four decades to overcome the limitations of static PRA methods. DPRA presents a comprehensive approach to assess the risks associated with complex, dynamic systems. However, current DPRA approaches are faced with challenges associated with the intra/interdependence within/between different NPP complex systems and the massive amount of data that needs to be analyzed and rapidly acted upon. In response to these limitations of previous work, the main objective of this dissertation is to develop a physics-based DPRA platform and an intelligent data-driven prediction tool for NPP safety enhancement under normal and abnormal operating conditions. The results of this dissertation demonstrate that the developed DPRA platform is capable of simulating the dynamic interaction between different NPP systems and estimating the temporal probability of core damage under different transients with significant analysis advantages from both the computational time and data storage perspectives. The developed platform can also explicitly account for uncertainties associated with the NPP's physical parameters and operating conditions on the plant's response and probability of its core damage. Furthermore, an intelligent decision support tool, developed based on artificial neural networks (ANN), can significantly improve the safety of NPPs by providing the plant operators with fast and accurate predictions that are specific to such NPP. Such rapid prediction will minimize the need to resort to idealized physics-based simulators to predict the underlying complex physical interactions. Moving forward, the developed ANN model can be trained under plant operational data, plants operating experience database, and data from rare event simulations to consider for example plant ageing with time, operational transients, and rare events in predicting the plant behavior. Such intelligent tool can be key for NPP operators and managers to take rapid and reliable actions under abnormal conditions. / Thesis / Doctor of Philosophy (PhD)

No description available.

The load forecasting has become an important role in the operation of power system, and several models by using different techniques have been applied to solve these problems. In the literature, the linear regression models are considered as a traditional approach to predict power consumption, and more recently, the artificial neural network (ANN) models have received more attention for a great number of successful and practical applications. This report introduces both linear regression and ANN models to predict the power consumption for Fortum in Ekerö. The characteristics of power consumption of different kinds of consumers are analyzed, together with the effects of weather parameters to power consumption. Further, based on the gained information, the numerical models of load forecasting are built and tested by the historical data. The predictions of power consumption are focus on three cases separately: total power consumption in one year, daily peak power consumption during winter and hourly power consumption. The processes of development of the models will be described, such as the choice of the variables, the transformations of the variables, the structure of the models and the training cases of ANN model. In addition, two linear regression models will be built according to the number of input variables. They are simple linear regression with one input variable and multiple linear regression with several input variables. Comparison between the linear regression and ANN models will be carried out. In the end, it finds out that the linear regression obtains better results for all the cases in Ekerö. Especially, the simple linear regression outperforms in prediction of total power consumption in one year, and the multiple linear regression is better in prediction of daily peak load during the winter.

Controlling external actions with the voice is something humans have tried to do for a long time. There are many different ways to implement a voice control system, and many of these applications require internet connections. Leaving the application area limited, as commercially available voice controllers have been stagnating behind due to the cost of developing and maintaining. In this project an artifact was created to work as an easy to use, generic, voice controller tool that allows the user to easily create different voice commands that can be implemented in many different applications and platforms. The user shall have no need of understanding or experience of voice controls in order to use and implement the voice controller.

The pandemic that struck the entire world 2020 caused by the SARS-CoV-2 (COVID-19) virus, will have an enormous interest for statistical and economical analytics for a long time. While the pandemic of 2020 is not the first that struck the entire world, it is the first pandemic in history where the data were gathered to this extent. Most countries have collected and shared its numbers of cases, tests and deaths related to the COVID-19 virus using different storage methods and different data types. Gaining quality data from the COVID-19 pandemic is a problem most countries had during the pandemic, since it is constantly changing not only for the current situation but also because past values have been altered when additional information has surfaced. The importance of having the latest data available for government officials to make an informed decision, leads to the usage of Business Intelligence tools and techniques for data gathering and aggregation being one way of solving the problem. One of the mostly used software to perform Business Intelligence is the Microsoft develop Power BI, designed to be a powerful visualizing and analysing tool, that could gather all data related to the COVID-19 pandemic into one application. The pandemic caused not only millions of deaths, but it also caused one of the largest drops on the stock market since the Great Recession of 2007. To determine if the deaths or other reasons directly caused the drop, the study modelled the volatility from index funds using Generalized Autoregressive Conditional Heteroscedasticity. One question often asked when talking of the COVID-19 virus, is how deadly the virus is. Analysing the effect the pandemic had on the mortality rate is one way of determining how the pandemic not only affected the mortality rate but also how deadly the virus is. The analysis of the mortality rate was preformed using Seasonal Artificial Neural Network. Forecasting deaths from the pandemic using the Seasonal Artificial Neural Network on the COVID-19 daily deaths data.

The purpose of this study was to discover factors about first-time freshmen that began at one of the six 4-year universities in the former Tennessee Board of Regents (TBR) system, transferred to any other institution after their first year, and graduated with a degree or certificate. These factors would be used with predictive models to identify these students prior to their initial departure. Thirty-four variables about students and the institutions that they attended and graduated from were used to perform principal component analysis to examine the factors involved in their decisions. A subset of 18 variables about these students in their first semester were used to perform principal component analysis and produce a set of 4 factors that were used in 5 predictive models. The 4 factors of students who transferred and graduated elsewhere were “Institutional Characteristics,” “Institution’s Focus on Academics,” “Student Aptitude,” and “Student Community.” These 4 factors were combined with the additional demographic variables of gender, race, residency, and initial institution to form a final dataset used in predictive modeling. The predictive models used were a logistic regression, decision tree, random forest, artificial neural network, and support vector machine. All models had predictive power beyond that of random chance. The logistic regression and support vector machine models had the most predictive power, followed by the artificial neural network, random forest, and decision tree models respectively.

[pt] O Brasil vem registrando a cada ano um crescimento expressivo no número de conexões de geração distribuída na rede de distribuição devido à concessão de incentivos governamentais que permitiu a difusão do uso de placas solares fotovoltaicas, fonte de geração de energia mais usada na geração distribuída no Brasil. Em sistemas elétricos com alta penetração de fontes intermitentes a previsão do comportamento da curva de carga tende a representar um grande desafio para os operadores do sistema devido à imprevisibilidade associada à geração de energia, podendo impactar diretamente no planejamento e operação da rede elétrica. Para lidar com esse desafio, este trabalho propõe uma metodologia de previsão de carga usando redes neurais recorrentes com arquitetura LSTM, considerando o impacto da mini e microgeração distribuída solar fotovoltaica conectada à rede de distribuição brasileira. São feitas previsões de carga do Sistema Interligado Nacional brasileiro e dos subsistemas que o integram, levando em conta um horizonte de curto prazo, de 24 horas, em intervalos horários, e um horizonte de médio prazo, de 60 meses, em intervalos mensais. Os resultados indicam que a metodologia pode ser uma ferramenta eficiente para a obtenção de previsões de carga podendo ser utilizada também para horizontes de previsão distintos dos apresentados neste trabalho. O MAPE encontrado para as previsões de curto prazo não passam de 2 por cento e para as previsões de médio prazo não passam de 3,5 por cento. / [en] Every year, Brazil has been registering a significant growth in the number of distributed generation connections in the distribution grid due to the granting of government incentives that allowed the use of solar photovoltaic panels to spread, the most used source of energy in distributed generation in Brazil. In electrical systems with high penetration of intermittent sources, the prediction of the behavior of the load curve tends to represent a great challenge for system operators due to the unpredictability associated with power generation, which can directly impact the planning and operation of the electrical grid. To deal with this challenge, this work proposes a load forecasting methodology using recurrent neural networks with LSTM architecture, considering the impact of the distributed photovoltaic solar generation connected to the Brazilian distribution grid. Load forecasts are made for the Brazilian National Interconnected System and for the subsystems that integrate it, taking into account a short-term horizon, of 24 hours, in hourly intervals, and a medium-term horizon, of 60 months, in monthly intervals. The results indicate that the methodology can be an efficient tool for obtaining load forecasts and can also be used for different forecast horizons than those presented in this work. The MAPE found for short-term forecasts is no more than 2 percent and for medium-term forecasts, no more than 3.5 percent.

This master thesis handles the development of an automatic benchmarking program for wind turbines and the thesis works as the theoretical basis for this program. The program is created at the request of the power company OX2 who wanted this potential to be investigated. The mission given by the company is to: 1. to find a good key point indicator for the efficiency of a wind turbine, 2. to find an efficient way to assess this and 3. to write a program that does this automatically and continuously. The thesis determines with a study of previous research that the best method to utilize for these kinds of continuous analyses are artificial neural networks which can train themselves on historical data and then assess if the wind turbine is working better or worse than it should with regards to its history. This comparison between the neural network predicted operation and the actual operation works as the measurement of the efficiency, the key point indicator for how the turbine work compared to how it historically should operate. The program is based on this principle and is completely written in MATLAB. Further testing of the program found that the best variables to use are wind speed and the blade pitch angle as input variables for the neural network and active power as the target used as the variable to predict and assess the operation. The final program was able to be fully automated and integrated into the OX2 system thanks to the possibility to continuously import wind turbine data through APIs. In the final testing was the program able to identify 75% of the anomalies manually found in the half year and in the five turbines used for this thesis, the small anomalies not found manually but identified by the program excluded. / Den här masteruppsatsen hanterar utvecklandet av ett automatiskt driftanalyseringsprogram för vindkraftverk och fungerar som det teoretiska underlaget för detta program. Programmet utvecklades på uppdrag av kraftbolaget OX2 som ville undersöka potentialen för ett sådant analysprogram i deras verksamhet. Uppdraget givet var att: 1. ta fram en bra indikator när det gäller den faktiska effektiviteten av ett vindkraftverk, 2. att hitta ett effektivt sätt att använda detta måttet i en analys där målet är att hitta avvikelser, och 3. skriva ett program som automatiskt kan använda måttet och metoden över tiden. Rapporten kommer via litteraturstudie fram till att tidigare forskning visar på att neurala nätverk är den mest lovande metoden för att genomföra sådan här analys. Dessa nätverk kan träna sig själva på historiska data och sedan analysera om vindturbinen arbetar bättre eller sämre än historiskt. Den här jämförelsen mellan den historiskt grundade förutspådda kraften ut och den faktiska kraften ut fungerar som kvalitetsmåttet på hur bra turbinen fungerar. Programmet är baserat på den här principen och är helt skriven i MATLAB. Vidare tester av programmet visar att de bästa variablerna att använda för att förutspå kraften ut är vindhastigheten och bladens vinkel mot vinden. Slutprogrammet var kapabelt att fullt automatiskt och integrerat i OX2s system identifiera 75% av alla avvikelser som manuellt hittats i ett halvårs data på de fem turbinerna använda för rapporten, småfel hittade av programmet men inte manuellt exkluderat.

Machine categorisation of the emotional content of music is an ongoing research area. Feature description and extraction for such a vague and subjective field as emotion presents a difficulty for human-designed audioprocessing. Research into machine categorisation of music based on genrehas expanded as media companies have increased their recommendation and automation efforts, but work into categorising music based on sentiment remains lacking. We took an informed experimental method towards finding a workable solution for a multimedia company, Ichigoichie, who wished to develop a generalizable classifier on musical qualities. This consisted of first orienting ourselves within the academic literature relevant on the subject, which suggested applying spectrographic pre-processing to the sound samples, and then analyzing these visually with a convolutional neural network. To verify this method, we prototyped the model in a high level framework utilizing Python which pre-processes 10 second audio files into spectrographs and then provides these as learning data to a convolutional neural network. This network is assessed on both its categorization accuracy and its generalizability to other data sets. Our results show that the method is justifiable as a technique for providing machine categorization of music based on genre, and even provides evidence that such a method is technically feasible for commercial applications today. / Maskinkategorisering av känsloprofilen i musik är ett pågående forskningsområde. Traditionellt sett görs detta med algoritmer som är skräddarsydda för en visstyp av musik och kategoriseringsområde. En nackdel med detta är att det inte går att applicera sådana algoritmer på flera användningsområden, och att det krävs både god musikkunnighet och även tekniskt vetande för att lyckas utveckla sådana algoritmer. På grund av dessa anledningar ökar stadigt mängden av forskning runt huruvida samma ändamål går att åstadkommas med hjälp av maskininlärningstekniker, och speciellt artificiella neuronnät, en delgrupp av maskininlärning. I detta forskningsprojekt ämnade vi att fortsätta med detta forskningsområde,och i slutändan hoppas kunna besvara frågan om huruvida det går att klassificera och kategorisera musik utifrån känsloprofilen inom musiken, med hjälp av artificiella neuronnät. Vi fann genom experimentell forskning att artificiella neuronnät är en mycket lovande teknik för klassificering av musik, och uppnådde goda resultat. Metoden som användes bestådde av spektrografisk ljudprocessering, och sedan analys av dessa spektrogram med konvolutionella neuronnät, en sorts artificiella neuronnät ämnade för visuell analys.