In this study, 22 homes in Milton, Ontario had their electricity consumption monitored for between seven and 15 months, and they were provided access to their data via an online webportal. The webportal provided appliance-level and house-level data, allowed them to set consumption goals, and schedule when their appliances would be used. The households were chosen to participate because they had previously expressed interest in advanced smart meter grid technologies, and when contacted again by Milton Hydro, they agreed to participate in the study. The main question being asked in this research is: what effect, if any, does having access to one???s consumption data have on consumption? To investigate this question, consumption data from the monitoring period, and the previous year (the base year) were obtained from Milton Hydro and were used to determine how consumption changed between these two periods. The consumption data for the cooling months were weather normalized to account for increases in consumption that result from cooling the dwelling. Data regarding users??? engagement with the webportal, including how often they would login, for how long and what pages they were visiting, were collected from the webportal. An engagement index was adapted and refined from Peterson & Carrabis (2008), and along with the engagement data from the webportal, was used to calculate the engagement index. Data from two surveys were used to profile the households and to investigate their attitudes and behaviours towards electricity consumption. There were several key findings. First, engagement with the webportal was quite low; the engagement index (a value between zero and one) for the first three months the hub was open averaged 0.285 and ranged from 0 to 0.523. These numbers dropped by the end of the seventh month to an average engagement index of 0.163, and ranged from 0 to 0.341. The second key finding was that the hubs were not consistently conserving electricity; for the first three months, 10 of the 22 households had conserved electricity between the base year and monitoring period; at the end of the seventh month, this dropped to nine households. At the end of the third month, the change in consumption was an increase of 8.22%, and at the end of month seven it was an increase of 7.71%. The third finding was that there did not appear to be a connection between energy conserving attitudes and energy conserving behaviours. In the surveys, 12 households stated that their goal was to conserve electricity, however, of these 12, only four actually conserved electricity at the end of month seven. Finally, when comparing the engagement index with the change in consumption, there appeared to be only a weak, negative correlation between the variables. This weak correlation may be a result of two things: (1) a lack of engagement, which limits the ability to find correlation between engagement and change in consumption; (2) there is actually a weak relation between the two variables. Based on these findings, some recommendations are put forth, specifically about how to engage householders with the webportal. Suggestions include getting applications for mobile devices, and delivering electricity saving tips to households via e-mail, text message, and/or on the homepage of the portal. These tips could be given based on the season, or based on the goals that were set, and would encourage and explain to householders how to decrease consumption.

Orientador: Prof. Dr. João Manoel Losada Moreira / Dissertação (mestrado) - Universidade Federal do ABC. Programa de Pós-Graduação em Energia, 2015. / O consumo residencial de energia elétrica no Brasil aumentou 64% nos últimos dez anos enquanto o consumo total de energia elétrica no País aumentou 51%. A intensidade elétrica dos domicílios definida, como a razão entre consumo de eletricidade domiciliar e o consumo efetivo das famílias, diminuiu 12% no período estudado, como também diminuiu a tarifa de eletricidade num 18%. Neste trabalho estuda-se o comportamento do consumo de eletricidade residencial em função dos fatores consumo efetivo das famílias, número de domicílios e tarifa de eletricidade. Duas técnicas foram utilizadas para a analise do consumo de eletricidade: a) análise de séries temporais para obter regressões do consumo de eletricidade em função de variáveis explicativas. A validade da regressão foi verificada por meio de testes de raiz unitária e de cointegração; b) técnica de decomposição LMDI ("logarithmic mean weight divisia method"). Os resultados da regressão linear forneceram elasticidades que permitiram avaliar e projetar no longo prazo o consumo de eletricidade. Os valores obtidos para as elasticidades para o período 1985-2013 foram 0,97 para o número de domicílios, 0,35 para consumo efetivo das famílias e - 0,32 para a tarifa. Os resultados mostram que o consumo de eletricidade apresenta mais sensibilidade às variações na variável numero de domicílios, isto é, o crescente aumento do número de residências no país é o responsável principal pelo aumento do consumo de eletricidade residencial. As variáveis explicativas consumo efetivo das famílias e tarifa de eletricidade variaram mais no período analisado que o número de domicílios, que apresenta um crescimento mais uniforme. Confirmou-se que a tarifa é uma possível variável controladora do consumo de eletricidade residencial por afetar indiretamente as preferências e hábitos das famílias. Para ser efetiva na redução de consumo de energia residencial ela deve ter uma taxa de variação maior que a taxa de variação do consumo efetivo das famílias, pois suas elasticidades são muito próximas, mas de sinais contrários. A partir da decomposição pela técnica LMDI, obteve-se a contribuição de cada variável explicativa no consumo de eletricidade, confirmando que a técnica é útil para conhecer e analisar os fatores em que a eletricidade decompõe-se, e não como uma técnica de projeção do consumo de eletricidade. / The residential electricity consumption in Brazil increased 64 % between 2003 and 2013 while the total electricity consumption in the country increased 51 %. The electric intensity of households, defined as the ratio of household consumption of electricity and the final consumption of households fell 12% during the study period, and the electricity tariff fell 18 %. In this work we study the residential electricity consumption behavior in terms of actual final consumption of household, number of households and electricity tariff. Two techniques were used for the analysis of electricity consumption: a) time-series analysis for regressions of electricity consumption in terms of explanatory variables. The validity of the regression was verified by unit root test and cointegration test; b) LMDI decomposition technique ("logarithmic mean weight dividing method"). The results of linear regression provided elasticities that allow us to evaluate and manage the long-term consumption of electricity. The values obtained for the elasticities for the period 1985-2013 were 0.97 for the number of households, 0.35 to actual final consumption of household and -0.32 for the electricity tariff. The results show that electricity consumption has more sensitivity to changes in the variable number of households, that is, the increasing number of households in the country is primarily responsible for the increase in residential electricity consumption. The explanatory variables consumption of household and electricity tariff varied over the analyzed period while the number of households presented a uniform growth. The electricity tariff may be used to manage the residential electricity demand. For reducing residential electricity consumption, its growth rate should be higher than that of the consumer spending because their elasticity¿s are very close, but of opposite signs. From the decomposition by LMDI technique, it obtained the contribution of each explanatory variable in electricity consumption, confirming this technique useful to know and analyze the factors on which electricity decomposes, and not as projection technique electricity consumption.

Conserving energy and reducing electricity consumption have become critical issues. Measuring when different appliances use electricity can be an effective way to save money on electricity bills. By providing information about hourly electricity prices and peak consumption times, people can subconsciously adopt energy saving habits to reduce the electricity consumption in their households. The challenge, however, lies in ensuring that all household members are informed and made aware of the right times to use electricity. This study will research how nudging can be utilized to reduce electricity consumption in households and what information the people in the households need to be able to make informed decisions to reduce their electricity consumption.

Orientador: Carlos Roberto Minussi / Resumo: Esta pesquisa será dedicada ao desenvolvimento de uma metodologia com vistas à compreensão e ao exame do comportamento do hábito de consumo de eletricidade residencial, via análise de similaridade, baseado no uso de uma rede neural da família ART (Adaptive Resonance Theory). Trata-se de uma rede neural composta por dois módulos ART-Fuzzy, cujo treinamento é realizado de modo não supervisionado. No primeiro módulo, serão usadas, como entrada, as informações que caracterizam os hábitos de consumo e a situação socioeconômica. A saída do primeiro módulo junto com os dados referentes aos equipamentos eletroeletrônicos da residência compõem a entrada do segundo módulo que, finalmente, produz informações, na saída, relativas ao diagnóstico pretendido, ou seja, a formação de agrupamentos similares (clusters). Todo o processamento da rede neural modular é realizado com dados binários, os quais são gerados a partir de informações quantitativas e qualitativas. As redes neurais da família ART são estáveis e plásticas. A estabilidade refere-se à garantia de sempre produzir soluções, ou seja, não se observa problemas relativos à má convergência. A plasticidade é uma característica que possibilita a execução do treinamento de forma contínua sem destruir o conhecimento adquirido previamente. É um recurso pouco observado nas demais redes neurais disponíveis na literatura especializada. Com essas propriedades (estabilidade e plasticidade), combinada com o processamento de dados essencialmente ... (Resumo completo, clicar acesso eletrônico abaixo) / Doutor

Submitted by Daniela Sbizera Justo null (sbizera@yahoo.com) on 2016-09-20T14:14:51Z No. of bitstreams: 1 Tese-Daniela Sbizera Justo.pdf: 5782774 bytes, checksum: 483d11758263a9d6c3a3d4c89fe66919 (MD5) / Approved for entry into archive by Ana Paula Grisoto (grisotoana@reitoria.unesp.br) on 2016-09-22T19:44:56Z (GMT) No. of bitstreams: 1 justo_ds_dr_ilha.pdf: 5782774 bytes, checksum: 483d11758263a9d6c3a3d4c89fe66919 (MD5) / Made available in DSpace on 2016-09-22T19:44:56Z (GMT). No. of bitstreams: 1 justo_ds_dr_ilha.pdf: 5782774 bytes, checksum: 483d11758263a9d6c3a3d4c89fe66919 (MD5) Previous issue date: 2016-08-25 / Esta pesquisa será dedicada ao desenvolvimento de uma metodologia com vistas à compreensão e ao exame do comportamento do hábito de consumo de eletricidade residencial, via análise de similaridade, baseado no uso de uma rede neural da família ART (Adaptive Resonance Theory). Trata-se de uma rede neural composta por dois módulos ART-Fuzzy, cujo treinamento é realizado de modo não supervisionado. No primeiro módulo, serão usadas, como entrada, as informações que caracterizam os hábitos de consumo e a situação socioeconômica. A saída do primeiro módulo junto com os dados referentes aos equipamentos eletroeletrônicos da residência compõem a entrada do segundo módulo que, finalmente, produz informações, na saída, relativas ao diagnóstico pretendido, ou seja, a formação de agrupamentos similares (clusters). Todo o processamento da rede neural modular é realizado com dados binários, os quais são gerados a partir de informações quantitativas e qualitativas. As redes neurais da família ART são estáveis e plásticas. A estabilidade refere-se à garantia de sempre produzir soluções, ou seja, não se observa problemas relativos à má convergência. A plasticidade é uma característica que possibilita a execução do treinamento de forma contínua sem destruir o conhecimento adquirido previamente. É um recurso pouco observado nas demais redes neurais disponíveis na literatura especializada. Com essas propriedades (estabilidade e plasticidade), combinada com o processamento de dados essencialmente binários, confere ao sistema neural uma ampla capacidade de produzir objetivos que podem ser facilmente modificados visando atender requisitos preestabelecidos pelos usuários (consumidor, empresa do setor elétrico). Neste sentido, o resultado esperado é a obtenção de informações referentes à similaridade de consumidores, à qual pode-se vislumbrar alguns benefícios, por parte dos consumidores, como melhorar o hábito de consumir energia elétrica, oferecendo também, por meio do conhecimento dos consumidores similares, a obtenção de melhores estratégias de negociação com os fornecedores, principalmente, no caso de sistemas smart grids. Neste novo paradigma do setor elétrico, há uma forte tendência do(s) consumidor(es) escolher(em) livremente a empresas fornecedoras de energia elétrica. Além disso, é discutida uma melhor forma para a realização da previsão de carga em pontos da rede elétrica onde há uma maior incerteza, e.g., nos barramentos mais próximos do consumidor (transformadores etc.), i.e., as incertezas no contexto da previsão de carga total do sistema são aumentadas à medida que se adentra a partir da carga global até chegar ao consumidor final, em especial ao usuário residencial. A base de dados, para a fase de treinamento da rede neural, é construída a partir de informações disponibilizadas por consumidores voluntários via o preenchimento de formulário. Realizada a fase de treinamento, a rede neural adquire um conhecimento incipiente afeito de ser aperfeiçoado ao longo do tempo, quando se implementa o recurso da plasticidade. / This work develops a methodology to understand and analyze the behavior of residential electricity consumption by similarity analysis, based on a neural network of ART (Adaptive Resonance Theory) family. The neural network is composed of two Fuzzy-ART modules whose training are non-supervised. At the first module, the inputs are information that characterize the consumption habits and the socio-economic situation. The output of the first module with the data referred to electro-electronic equipment available at the residence compose the input of the second module, which finally produces information at the output related to the diagnosis proposed, i.e. the formation of clusters. All the neural network processing is realized with binary data, which are generated from quantitative and qualitative information. ART family neural networks are stable and plastic. The stability assures that it always produces a solution, i.e. there is no convergence problem. The plasticity is a characteristic that allows executing the processing continuously without losing the knowledge previously learned. Those advantages are seldom observed in other neural networks available at the specialized literature. Considering these properties (stability and plasticity), combined with the data processing exclusively binary, the neural network is capable to be modified when necessary to attend pre-defined requests by the users (consumers, distributers, etc.). Therefore, the expected result is to obtain information referred to the similarity with consumers, and with this information, the consumers can improve their habits or even negotiating with the producers in case of smart grid systems. This new electrical system paradigm, the tendency is that the consumers can arbitrarily choose the electrical distributers. Furthermore, the work discusses the best way to realize load forecasting in points where there is uncertainty, e.g., on the busses near the consumers (transformers), i.e., the uncertainties considering the global forecasting increase if the information of residences is not considered. The database for the training phase of the neural network was built by a quiz form filled by some volunteer consumers. Afterwards, when finishing the training phase, the neural network acquires knowledge that along time can implement the plasticity resource.

Abstract Achieving the objective of a decarbonised economy by 2050 will require massive efforts in the energy sector. Emissions from residential houses will have to be almost completely cut, by around 90% by 2050. Home automation is a potential tool for achieving this goal. However, the environmental and economic benefits of automation technologies first need to be assessed. This thesis evaluates the impact of home automation for electricity management in the residential sector using environmental and economic indicators. To this end, a life cycle assessment was performed to evaluate the impacts of the manufacturing, use and disposal phases. The influences of end-user behaviour, household size and multiple levels of technological deployment were also investigated. A Markov chain simulation tool, built on the MatLab platform, was developed to assess all possible combinations of impacting factors. Dynamic environmental indicators were developed based on the ReCiPe method for aggregating the impacts of processes. All these indicators were then combined to form a single index based on multi-criteria acceptability analysis. The results suggest that home automation can decrease peak load, but that overall electricity consumption may increase due to electricity use by the actual automation system. The effect of home automation was more noticeable in larger households than in one-person households. In addition, use of dynamic environmental indicators proved more relevant than fixed indicators to represent the environmental impact of home automation. Within the life cycle of automation technology, the manufacturing phase had the highest impact, but most of the CO2 emissions originated from the use phase. In conclusion, the most important environmental benefit of home automation is reducing CO2 emissions during peak time by load shifting. / Tiivistelmä Vähähiilisen talouden saavuttaminen vuoteen 2050 mennessä edellyttää valtavia ponnisteluja energia-alalla. Rakennuksista aiheutuvia päästöjä on vähennettävä radikaalisti, jopa 90 % vuoteen 2050 mennessä. Rakennusten energiatehokkuutta edistävä automaatiotekniikka on yksi keino tämän päämäärän saavuttamiseen. Kotiautomaation kautta voidaan sekä vähentää energian kokonaiskulutusta että tasoittaa energiankäyttöprofiilia. On kuitenkin tutkittava myös, mitkä ovat automaatiotekniikan ympäristö- ja taloudelliset vaikutukset. Tässä työssä käsitellään kotiautomaation vaikutusta sähkön kulutuksen hallintaan asuinrakennuksissa käyttämällä ympäristö- ja talousindikaattoreita. Tätä varten suoritettiin kotiautomaation elinkaariarviointi selvittämällä laitteiden valmistus-, käyttö- ja hävittämisvaiheiden ympäristövaikutukset. Työssä tarkasteltiin myös asukkaiden käyttäytymisen, kotitalouden koon ja eri teknologiavaihtoehtojen vaikutuksia ympäristö- ja talousvaikutuksiin. Arviointi suoritettiin Markovin ketjun simulointityökalulla, joka rakennettiin Matlab-alustalle. Dynaamisia ympäristömittareita kehitettiin ReCiPe-menetelmää käyttäen. Indikaattorit on edelleen yhdistetty yhdeksi indeksiksi käyttäen monikriteeriarviointia. Tulokset viittaavat siihen, että huippukuormitusta voidaan vähentää käyttämällä kotiautomaatiota, mutta sähkön kokonaiskulutus voi kasvaa automaatiojärjestelmän sähkönkulutuksen takia. Kotiautomaation vaikutukset ovat eniten havaittavissa suurissa kotitalouksissa. Lisäksi, dynaamiset indikaattorit edustavat paremmin kotiautomaation vaikutusta ympäristöön kuin staattiset indikaattorit. Automaatioteknologian elinkaaressa suurimmat ympäristövaikutukset ovat valmistusvaiheessa, mutta CO2-päästöjä syntyy eniten käyttövaiheessa. Lopuksi voidaan todeta, että kotiautomaation merkittävin ympäristöhyöty on CO2-päästöjen vähentäminen huippukulutuksen aikana siirtämällä kuormitusta toiseen ajankohtaan.