FundaÃÃo Cearense de Apoio ao Desenvolvimento Cientifico e TecnolÃgico / O monitoramento do carregamento de componentes elÃtricos constitui-se em um aspecto de grande importÃncia para qualquer sistema de potÃncia, pois a partir dele podem ser observadas as condiÃÃes de seguranÃa dos seus componentes. O presente trabalho propÃe um mÃtodo computacional baseado em Redes Neurais Artificiais para monitorar o carregamento de componentes elÃtricos do sistema de distribuiÃÃo de energia, como transformadores, alimentadores e linhas de subtransmissÃo. A partir dos dados colhidos dos medidores das subestaÃÃes, contendo os valores de corrente do componente considerado, sÃo realizadas duas anÃlises: com transferÃncia de carga e sem transferÃncia de carga. Desta maneira, objetiva-se determinar os valores mÃximos de corrente nas duas situaÃÃes para o correspondente transformador, alimentador, ou linha de subtransmissÃo analisada. Busca-se entÃo, obter o seu carregamento mÃximo em ambos os casos e a partir desses valores, determinar se o componente està ou nÃo operando em boas condiÃÃes de seguranÃa. O valor mÃximo de corrente com transferÃncia de carga à simples de ser obtido, pois consiste apenas no valor mÃximo dos dados de corrente sem qualquer tipo de anÃlise mais aprofundada. PorÃm, o valor mÃximo de corrente sem transferÃncia de carga à bastante complexo de ser determinado, pois as condiÃÃes atÃpicas dos dados devem ser eliminadas. Desta forma, um mÃtodo empregando Redes Neurais Artificiais foi desenvolvido para obter este valor de corrente para os componentes analisados. Os resultados se mostraram bem prÃximos dos valores reais, comprovando a eficÃcia do mÃtodo. Finalmente, pode ser concluÃdo que o monitoramento à perfeitamente possÃvel de ser realizado, possibilitando um maior controle sobre os carregamentos, evitando danos tanto ao sistema de distribuiÃÃo, como ao sistema de potÃncia como um todo. / The monitoring of the electrical components loading is an aspect of great importance to any power system, since the components safety conditions can be observed from it. The present study proposes a computational method based on Artificial Neural Networks to monitor the electrical components loading of the power distribution system, as transformers, feeders and subtransmission lines. Based on the data collected from the substations meters, containing the electric current values of the considered component, two analyses are done: with load transfer and without load transfer. Thus, the aim is to determine the current maximum values in both situations to the corresponding transformer, feeder, or subtransmission line analyzed. Then it is sought to obtain the maximum loading in both cases and, from these values, to determinate whether or not the component is operating in good safety conditions. The maximum current with load transfer is simple to obtain, because it consists only in the maximum value of electrical current data without any deeper analysis. However, the maximum current without load transfer is very complex to be determined, once the atypical conditions of the data must be eliminated. Thereby, a method using Artificial Neural Networks was developed to estimate the values of the current to the analyzed components. The results were very close to the real ones, proving the effectiveness of the method. Finally, it can be concluded that the monitoring is perfectly possible to be performed, allowing greater control over the loadings, avoiding damages to both the distribution system and the power system as a whole.

The breakthrough in mobile technology and the development of smartphones, supplied with sensing devices such as Inertial Measurement Units (IMUs), has made it possible to obtain accurate and reliable data on the angular velocity for different objects. The available technical sensors for wrist movements, such as electrogoniometers, are costly, time-consuming, and need a particular computer program to be analyzed. Therefore, there is a need to develop user-friendly risk assessment methods for wrist angular velocity measurements. This master thesis aimed to validate the accuracy of a newly developed iPhone application (App), "ErgoHandMeter," for wrist velocity in actual work tasks, by comparing the “ErgoHandMeter” to standard electrogoniometers. The project study was performed with four participants, two females and two males, from three jobs performing actual work tasks. The total angular velocity obtained by the mobile application was compared with the angular velocity data from the standard electrogoniometer. The total angular velocities obtained from the smartphone and the goniometer were computed at the 10th, 50th and 90th percentile for the four subjects. The 50th percentile of goniometer-flexion velocity (G-flex) was 7.4 ± 5.4°/s, for the goniometer-total (G-tot) 8.7 ± 6.5)°/s and for App 7.2 ± 4.9°/s. The correlation coefficient for the 50th percentile of goniometer-flexion (G-flex) parameter and smartphone application was 0.994. For the goniometer-total (G-tot) and the application, it was 0.993. In a Bland-Altman plot the mean difference between G-flex and App for the 50th percentile was -0.18 °/s and for G-tot and App was -1.54 °/s, i.e. the App was lower in average. The limit of the agreement between G-Flex and App, and G-tot and App stayed within two standard deviations. For G-Flex and App (mean+1.96SD) was 1.34 °/s, (mean-1.96SD) was -1.71 °/s, while for G-tot and App (mean+1.96SD) was 1.89 °/s, (mean-1.96SD) was -4.96 °/s, indicating an adequate agreement between the two methods. A limitation was that the included occupations were all relatively low velocity. However, in conclusion, the results indicate that the two methods agree adequately and can be used interchangeably. / Genombrottet inom mobiltekniken och utvecklingen av smarttelefoner med sensorer som t.ex. tröghetsmätningsenheter (IMU) har gjort det möjligt att få exakta och tillförlitliga uppgifter om vinkelhastigheten för olika objekt. De tillgängliga tekniska sensorerna för handledsrörelser, t.ex. elektrogoniometrar, är dyra, tidskrävande och de samplade signalerna kräver ett särskilt datorprogram för att analyseras. Det finns därför ett behov av att utveckla användarvänliga riskbedömningsmetoder för mätningar av handledens vinkelhastighet. Syftet med detta examensarbete var att validera noggrannheten hos en nyutvecklad iPhone-applikation (App), "ErgoHandMeter", för handledshastighet i verkliga arbetsuppgifter, genom att jämföra "ErgoHandMeter" med vanliga elektrogoniometrar. Projektstudien genomfördes med fyra deltagare, två kvinnor och två män, från tre yrken som utförde verkliga arbetsuppgifter. Den totala vinkelhastigheten som erhölls av mobilapplikationen jämfördes med vinkelhastighetsdata från standardelektrogoniometern. De totala vinkelhastigheterna som erhållits från smarttelefonen och goniometern beräknades vid den 10:e, 50:e och 90:e percentilen för de fyra försökspersonerna. Den 50:e percentilen för goniometer-flexionshastigheten (G-flex) var i genomsnitt 7,4°/s och för goniometertotalen (G-tot) 8,7°/s. Korrelationskoefficienten (r) för den 50:e percentilen för goniometer-flexionsparametern (G-flex) och smartphone-applikationen var 0,994. För goniometer-total (G-tot) och applikationen var r 0,993. I en Bland-Altman-plot var den genomsnittliga skillnaden mellan G-flex och appen för den 50:e percentilen -0,18°/s och för G-tot och appen -1,54°/s (App var lägre än Gon). Medelvärdet för differensen mellan G-Flex och App och G-tot och App ligger inom två standardavvikelser. För G-Flex och App (medelvärde+1,96SD) var 1,34 °/s, (medelvärde-1,96SD) var -1,71 °/s, medan för G-tot och App (medelvärde+1,96SD) var 1,89 °/s, (medelvärde-1,96SD) var -4,96 °/s. Vilket tyder på en tillräcklig överensstämmelse mellan de två metoderna. En begränsning var att de inkluderade yrkena alla hade relativt låg hastighet. Sammanfattningsvis visar dock resultaten att de två metoderna stämmer väl överens och kan användas på ett utbytbart sätt.