Non-technical losses (NTL) have been a growing problem over the years, causing significant financial losses for electric utilities. Among the methods for detecting this type of loss, those based on Artificial Intelligence (AI) have been the most popular. Many works use the electricity consumption profile as an input for AI models, which may not be sufficient to develop a model that achieves a high detection rate for various types of energy fraud that may occur. In this paper, using actual electricity consumption data, additional statistical and temporal features based on these data are used to improve the detection rate of various types of NTL. Furthermore, a model that combines both the electricity consumption data and these features is developed, achieving a better detection rate for all types of fraud considered.

In this chapter it is elucidated how the analytical solution was obtained to obtain the electromagnetic polarization profile of a completely polarized wave in the region of distant fields. The analytical solution was obtained from the interpretation of the physical phenomenon associated with the method of the linear component, which was adapted for the use of discrete and miniaturized elements, reducing the physical space for the measurement circuit and the expenses associated to the delay circuit. From the analytical solution it is possible to observe that with only one phase measurement in the delay circuit it is possible to obtain the polarization profile of the wave, with the axial ratio and phase, respectively.

This research aims to evaluate the efficiency of thermography as a tool for the detection of faults in low voltage electrical networks. For this, a thermovisor was used in several equipment, such as panels, transformers and capacitor banks, to locate possible thermal anomalies. Two analyzes were made: one qualitative and the other quantitative. The first is based on the investigation between an adjacent element and the suspicious element and a definition of criticality, in which actions are taken to correct failures. The quantitative analysis aims at improving a definition of criticality where possible, such as not identifying the thermal emissivity of the materials involved that are not identified by the electric currents of the circuits. In addition, for the cases studied, a criticality comparison was made between the two methods.

Diante do crescimento do uso da energia solar fotovoltaica (FV) no Brasil nos últimos anos, é importante poder mensurar quanto as usinas gerarão nas próximas horas ou dias, para que seja feito um planejamento do setor elétrico de forma confiável, garantindo a qualidade e o fornecimento contínuo de energia elétrica. Sendo assim, esse trabalho consistiu em determinar os melhores hiperparâmetros para a construção de um modelo de previsão de energia solar FV utilizando a rede neural recorrente Long Short-Term Memory (LSTM). Após a construção do modelo, foram realizados testes com os dados históricos de geração da usina solar de Pirapora – MG, com o objetivo de fazer a previsão dos valores de geração de forma diária e horária. Utilizando as métricas de erro, Raiz do Erro Médio Quadrático (RMSE), Erro Médio Absoluto (MAE) e Erro Médio Absoluto Percentual (MAPE), foram encontrados valores de 222,62 MWh, 153,85 MWh e 12,04% para o teste de forma diária e 41,21 MWh, 27,50 MWh e 36,07% no teste de forma horária, em que, para ambos os casos o período de teste foi de 6 meses. Além disso, também foram obtidos os valores do RMSE, MAE e MAPE para as primeiras 120 horas do teste, que foram 19,47 MWh, 13,83 MWh e 16,05% para o período em questão, respectivamente. Os resultados encontrados para as duas formas de previsão foram considerados eficientes, visando que foi utilizada apenas uma entrada na rede neural.