Despite the importance of power systems in today’s societies, they suffer from aging infrastructure and need to improve the efficiency, reliability, and security. Two issues that significantly limit the current grid’s efficient energy delivery and consumption are: loadollowing generation dispatch, and energy theft. A loadollowing generation dispatch is usually employed in power systems, which makes continuous small changes so as to account for differences between the actual energy demand and the predicted values. This approach has led to an average utilization of energy generation capacity below 55% [49]. Moreover, energy theft causes several billion dollar losses to U.S. utility companies [31] [16], while in developing countries it can amount to 50% of the total energy delivered [48]. Recently, the Smart Grid has been proposed as a new electric grid to modernize current power grids and enhance its efficiency, reliability, and sustainability. Particularly, in the Smart Grid, a digital communication network is deployed to enable two-way communications between users and system operators. It thus makes it possible to shape the users’ load demand curves by means of demand response strategies. Additionally, in the Smart Grid, traditional meters will be replaced with cyber-physical devices, called smart meters, capable of recording and transmitting users’ real-time power consumption. Due to their monitoring capabilities, smart meters offer a great opportunity to detect energy theft in smart grids, but also raise serious concerns about users’ privacy. In this dissertation, we design optimal load scheduling schemes to enhance system efficiency, and develop energy theft detection algorithms that can preserve users’ privacy.

[pt] As perdas de energia elétrica, particularmente as comerciais, constituem problema importante e complexo para as empresas distribuidoras. O problema só se apresenta de forma aguda nos países em desenvolvimento, onde o percentual de perdas por furto é elevado e adquire importantes conotações sociais, políticas e culturais. Isso explica o fato de empresas brasileiras de distribuição de energia elétrica investirem recursos apreciáveis para mitigar esse problema e ressalta a importância de um tema ainda pouco explorado na literatura técnico-científica. Esta dissertação, de caráter exploratório, tem como objetivo estudar o problema de perda de energia por desvio e a eficácia de soluções tecnológicas investigando o caso de uma empresa e os resultados obtidos em projeto recente. Ao tempo do estudo, a empresa implantava uma tecnologia cujo objetivo principal era identificar os clientes que desviam energia elétrica e, posteriormente, tornar o processo de faturamento automatizado. O estudo partiu de algumas questões julgadas mais relevantes, incluindo: Quanto a empresa ganhou com a instalação desta tecnologia? Como se comporta o consumo dos clientes antes e depois da instalação da tecnologia? Há correlação entre o ganho da empresa e o poder aquisitivo da população da área? Há efeito demonstração de uma área sobre suas vizinhas? A partir destas questões foram analisadas séries de dados referentes ao consumo medido antes e após a instalação da nova tecnologia. Os resultados indicam a viabilidade da tecnologia como eficaz paliativo de curto-prazo para o problema cuja solução definitiva envolve tratamento concomitante dos aspectos sociais, políticos e culturais. / [en] Electrical energy losses, chiefly, commercial losses, constitute an important and complex problem for energy distribution companies worldwide. However, only in developing countries the problem becomes acute because theft losses are large, and coupled with social, legal, political and cultural issues. This explains why Brazilian energy distribution companies dedicate sizable investments for mitigating the problem, and stresses the importance of a theme still largely unexplored in the technical-scientific literature. This thesis, of exploratory nature, addresses the problem and assesses the effectiveness of technological solutions by investigating the case of a recent project developed in a Brazilian company. At the time of the research, the company was implementing a new and original technology, whose main objective was to identify pilfering clients, and later on, to make the billing process more automated. The study started with some questions judged more relevant, including: How large was the gain yielded by the new technology? How does the energy consumption behave after the technological change? Is there a correlation between the gain obtained and the areas of lesser purchasing power? Is there a demonstration effect upon neighboring areas? Starting from these questions, series of historical data on measured consumption before and after the new technology were analyzed. The results indicate the feasibility of the technology as an effective shortterm palliative for the problem whose ultimate solution involves the concurrent treatment of the social, legal, political e cultural aspects.

Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-01-03T12:10:09Z No. of bitstreams: 1 matheusalbertodesouza.pdf: 12016608 bytes, checksum: 7403ebccb4f05e67a5f65a371f64f7e4 (MD5) / Approved for entry into archive by Diamantino Mayra (mayra.diamantino@ufjf.edu.br) on 2017-01-31T10:31:30Z (GMT) No. of bitstreams: 1 matheusalbertodesouza.pdf: 12016608 bytes, checksum: 7403ebccb4f05e67a5f65a371f64f7e4 (MD5) / Made available in DSpace on 2017-01-31T10:31:30Z (GMT). No. of bitstreams: 1 matheusalbertodesouza.pdf: 12016608 bytes, checksum: 7403ebccb4f05e67a5f65a371f64f7e4 (MD5) Previous issue date: 2016-09-30 / O presente trabalho tem por objetivo apresentar uma metodologia para a detecção de perdas comerciais de energia em sistemas de distribuição, bem como a identificação do usuário fraudulento em Infraestruturas Avançadas de Medição, próprias de Smart Grids. ParaaetapadedetecçãodeperdascomerciaisdeenergiaéutilizadoumEstimadorEstático de Estados Trifásico baseado em Medições Fasoriais Sincronizadas (Phasor Measurements Units - PMUs). Essa etapa visa detectar os transformadores na rede de distribuição, os quais, apresentam indícios de furtos/fraudes de energia, através da observância da curva de carga estimada. A próxima etapa visa identificar os consumidores, conectados aos transformadores suspeitos, que cometem o ato ilícito. Para isto, treinou-se uma Rede Neural Artificial (RNA) de Kohonen para clusterizar os consumidores honestos de acordo compadrõesdeconsumosemelhantes,apartirdedadosdeconsumoprovenientesdosSmart Meters decadacliente. ParacadaclassedefinidapelaRNAdeKohonencriou-seumaRNA MultiLayer Perceptron (MLP) para fins de classificação dos consumidores em duas classes, honestos ou fraudulentos. As duas grandes contribuições do trabalho estão em utilizar a técnica de detecção para diminuir o número de Falsos Positivos (FP) das metodologias de classificação, sem a necessidade de medições individuais nos transformadores, e a utilização de RNAs a partir de dados de medições do sistema de distribuição de forma robusta, em que não há a necessidade de medições sem perdas comerciais de energia para todos os consumidores do sistema. A forma de tratamento dos dados de medições ainda possibilita manter a privacidade dos usuários, questão bastante debatida no cenário mundial. Foram realizados testes para a etapa de detecção de perdas comerciais com os sistemas IEEE 33 Barras e IEEE 70 barras. Para a etapa de identificação do consumidor fraudulento foram utilizados dados de consumo reais de mais de 5000 consumidores provenientes de Smart Meters liberados pela Autoridade de Eletricidade e Energia Sustentável da Irlanda. Os testes mostraram bons resultados tornando a metodologia proposta aplicável na detecção e identificação de perdas comerciais de energia elétrica em Smart Grids. / This work aims to present a methodology for the detection of energy theft in distribution systems as well as the identification of the fraudulent users considering Advanced Metering Infrastructure, widely used in Smart Grids. For the stage of energy theft detection, a Static Three Phase State Estimator based on Synchronised Phasor Measurement Units (PMUs) is used. This step aims to detect the transformers in the distribution network, which have evidence of energy theft, by observing the estimated load curve. The next step is to identify consumers, connected to suspected transformers, which are stealing energy. For this, a Kohonen Artificial Neural Network (ANN) was trained to clustering honest consumers according to similar patterns of consumption, with the consumption data from the Smart Meters at every customer. For each class defined by the Kohonen ANN a MultiLayer Perceptron (MLP) ANN for classification of consumers into two classes, honest or fraudulent was created. The two major contributions of this work are the use of energy theft detection technique to reduce the number of false positives (FP) from the classification methods, without the need for individual measurements on the transformers, andtheuseofANNfromdistributionsystemmeasurementsmakethetechiniquerobust, in which there is no need for measurements without comercial loss of energy for all consumers. The way that the measurement data is treated allows maintaining the privacy of the consumers which is a debated question on the world. Tests were conducted for energy theft detection step with IEEE 33 Buses and IEEE 70 Buses systems. For the fraudulent user identification step, actual consumption data were used over 5000 consumers from their SmartMetersreleasedbytheElectricityAuthorityandSustainableEnergyofIreland. The tests showed good results making the proposed methodology applicable in the detection and identification of energy theft in Smart Grids.