Context: In recent years, households have been increasing energy consumption to very high levels, where it is no longer sustainable. There has been a dire need to find a way to use energy more sustainably due to the increase in the usage of energy consumption. One of the main causes of this unsustainable usage of energy consumption is that the user is not much acquainted with the energy consumed by the smart appliances (dishwasher, refrigerator, washing machine etc) in their households. By letting the household users know the energy usage consumed by the smart appliances. For the energy analytics companies, they must analyze the energy consumed by the smart appliances present in a house. To achieve this Kelly et. al. [7] have performed the task of energy disaggregation by using deep neural networks and producing good results. Zhang et. al. [7] has gone even a step further in improving the deep neural networks proposed by Kelly et. al., The task was performed by Non-intrusive load monitoring (NILM) technique. Objectives: The thesis aims to assess the performance of the deep neural networks which are proposed by Kelly et.al. [7], and Zhang et. al. [8]. We use deep neural networks for disaggregation of the dishwasher energy consumption, in the presence of vampire loads such as electric heaters, in a Swedish household setting. We also try to identify the training time of the proposed deep neural networks.  Methods: An intensive literature review is done to identify state-of-the-art deep neural network techniques used for energy disaggregation.  All the experiments are being performed on the dataset provided by the energy analytics company Eliq AB. The data is collected from 4 households in Sweden. All the households consist of vampire loads, an electrical heater, whose power consumption can be seen in the main power sensor. A separate smart plug is used to collect the dishwasher power consumption data. Each algorithm training is done on 2 houses with data provided by all the houses except two, which will be used for testing. The metrics used for analyzing the algorithms are Accuracy, Recall, Precision, Root mean square error (RMSE), and F1 measure. These software metrics would help us identify the best suitable algorithm for the disaggregation of dishwasher energy in our case.  Results: The results from our study have proved that Gated recurrent unit (GRU) performed best when compared to the other neural networks in our study like Simple recurrent neural network (SRN), Convolutional Neural Network (CNN), Long short-Term memory (LSTM) and Recurrent convolution neural network (RCNN). The Accuracy, RMSE and the F1 score of the GRU algorithm are higher when compared with the other algorithms. Also, if the user does not consider F1 score and RMSE as an evaluation metric and considers training time as his or her metric, then Simple recurrent neural network outperforms all the other neural nets with an average training time of 19.34 minutes.

This dissertation explores the problem of energy estimation in supervised Non-Intrusive Load Monitoring (NILM). NILM refers to a set of techniques used to estimate the electricity consumed by individual loads in a building from measurements of the total electrical consumption. Most commonly, NILM works by first attributing any significant change in the total power consumption (also known as an event) to a specific load and subsequently using these attributions (i.e. the labels for the events) to estimate energy for each load. For this last step, most proposed solutions in the field impart simplifying assumptions to make the problem more tractable. This has severely limited the practicality of the proposed solutions. To address this knowledge gap, we present a framework for creating appliance models based on classification labels and aggregate power measurements that can help relax many of these assumptions. Within the framework, we model the problem of utilizing a sequence of event labels to generate energy estimates as a broader class of problems that has two major components (i) With the understanding that the labels arise from a process with distinct states and state transitions, we estimate the underlying Finite State Machine (FSM) model that most likely generated the observed sequence (ii) We allow for the observed sequence to have errors, and present an error correction algorithm to detect and correct them. We test the framework on data from 43 appliances collected from 19 houses and find that it improves errors in energy estimates when compared to the case with no correction in 19 appliances by a factor of 50, leaves 17 appliances unchanged, and negatively impacts 6 appliances by a factor of 1.4. This approach of utilizing event sequences to estimate energy has implications in virtual metering of appliances as well. In a case study, we utilize this framework in order to substitute the need of plug-level sensors with cheap and easily deployable contacless sensors, and find that on the 6 appliances virtually metered using magnetic field sensors, the inferred energy values have an average error of 10:9%.

Le domaine lié à ce travail est appelé « désagrégation d’énergie », où la principale préoccupation est de décomposer, ou désagréger, la consommation globale d’énergie électrique (par exemple, la consommation de tout un ménage) en une consommation détaillée donnée comme information de consommation par usage (par exemple, par appareil). Cette dernière permet d’avoir un retour sur la consommation pour les consommateurs ainsi que pour les fournisseurs et est utile pour permettre des économies d’énergie. Dans ce domaine de désagrégation d’énergie, il existe trois grandes questions auxquelles il faut répondre : qui consomme ? quand ? et combien ? Les recherches menées dans cette thèse se concentrent sur l’identification des appareils électriques, c’est-à-dire la réponse à la première question, en considérant particulièrement des appareils ménagers. À cet effet, nous utilisons le courant transitoire de mise en marche que nous modélisons en utilisant un nouveau modèle que nous avons proposé. De plus, nous utilisons les paramètres estimés de ce dernier pour la tâche d’identification. / The related field to this work is called “energy disaggregation" where the main concern is to break down, or disaggregate, the global electrical energy consumption (e.g. wholehouse consumption) into a detailed consumption given as end-use (e.g. appliance-level) consumption information. This latter gives consumption feedback to consumers and electricity providers and is helpful for energy savings. Three main questions have to be answered in the energy disaggregation field : who is consuming ? when ? and how much ? The research conducted in this thesis focuses on electrical appliances identification, i.e. the who question, considering particularly home appliances. For this purpose, we use the turn-on transient current signal which we model using a new model we proposed and use its estimated model parameters for the identification task.

Submitted by Kelren Mota (kelrenlima@ufpa.br) on 2018-06-25T18:48:12Z No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Dissertacao_RedeNeuralConvolucional.pdf: 2088560 bytes, checksum: 6328f6f59bc552055a366b1e4a32793d (MD5) / Approved for entry into archive by Kelren Mota (kelrenlima@ufpa.br) on 2018-06-25T18:48:32Z (GMT) No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Dissertacao_RedeNeuralConvolucional.pdf: 2088560 bytes, checksum: 6328f6f59bc552055a366b1e4a32793d (MD5) / Made available in DSpace on 2018-06-25T18:48:32Z (GMT). No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Dissertacao_RedeNeuralConvolucional.pdf: 2088560 bytes, checksum: 6328f6f59bc552055a366b1e4a32793d (MD5) Previous issue date: 2018-04-03 / Este trabalho apresenta a proposta de uma nova metodologia para identificação de equipamentos residenciais em sistemas de Monitoramento Não-Intrusivo de cargas. O sistema é baseado em uma Rede Neural Convolucional para classificação dos equipamentos, que utilizam, diretamente como entradas para o sistema, os dados do sinal transitório de potência de 7 equipamentos obtidos no momento em que estes são ligados em uma residência. A metodologia foi desenvolvida usando dados de um banco de dados público (REED) que apresenta dados coletados a uma baixa frequência (1 Hz). Os resultados obtidos na base de dados de testes apresentam acurácia superior a 90%, indicando que o sistema proposto é capaz de realizar a tarefa de identificação, além disso os resultados apresentados são considerados satisfatórios quando comparados com os resultados já apresentados na literatura para o problema em questão. / This research presents the proposal of a new methodology for the identification of residential equipment in non-intrusive load monitoring systems. The system is based on a Convolutional Neural Network to classify residential equipment, which uses directly as inputs to the system, the transient power signal data of 7 equipment obtained at the moment they are connected in a residence. The methodology was developed using data from a public database (REED) that presents data collected at a low frequency (1 Hz). The results obtained in the test database show an accuracy of more than 90%, indicating that the proposed system is capable of performing the task of identification. In addition, the results presented are considered satisfactory when compared with the results already presented in the literature for the problem in question.

Comme les gens vieillissent, ils sont souvent confrontés à un certain degré de diminution des capacités cognitives ou de la force physique. Isolement de la vie sociale, mauvaise qualité de la vie, et risque accru de blessures en sont les principales conséquences. Ambient Assisted Living (AAL) est une vision de la façon dont les gens vivent leur vie dans leur propre maison, à mesure qu'ils vieillissent : handicaps ou limitations sont compensées par la technologie, là où le personnel de prestation de soins est rare ou des proches ne sont pas en mesure d'aider. Les personnes concernées sont assistés par la technologie. Le terme "ambiante" en AAL exprime, ce que cette technologie doit être, au- delà de l’assistance. Elle doit être intégrée dans l’environnement de manière à ce qu'elle ne soit pas reconnue en tant que tel. L'interaction avec les résidents doit être intuitive et naturelle. L'équipement technique doit être discret ct bien intégré. Les domaines d'application ciblés dans cette thèse sont le suivi de l’activité et la recherche de profils d'activités dans des appartements ou des petites maisons. L'acquisition d’informations concernant l’activité des résidents est vitale pour le succès de toute la technologie d’assistance. Dans de nombreux domaines de la vie quotidienne, ceci est déjà de la routine. L’état de l’art en matière de technologie de détection comprend des caméras, des barrières lumineuses, des capteurs RFID, la radiolocalisation de signal en utilisant des transpondeurs et des planchers sensibles à la pression. En raison de leurs principes de fonctionnement, ils ont malheureusement un impact important sur les environnements domestiques et de vie. Par conséquent, cette thèse est consacrée à la recherche de technologies d’acquisition d’informations de l’activité non-intrusive ayant un impact minimal sur la vie quotidienne. Deux technologies de base, la détection de présence passive sans dispositif et le suivi de charges de manière non-intrusive, sont prises en compte dans cette thèse. / As people grow older, they are often faced with some degree of decreasing cognitive abilities or physical strength. Isolation from social life, poor quality of life, and increased risk or injuries are the consequence. Ambient Assisted Living (AAL) is a vision for the way people live their life in their own home, as they grow older: disabilities or limitations are compensated for by technology, where care-giving personnel is scarce or relatives are unable to help. Affected people are assisted by technology. The term "Ambient" in AAL expresses, what this technology needs to be, beyond assistive. It needs to integrate into the living environment in such a way that it is not recognized as such any more. Interaction with residents needs to be intuitive and natural. Technical equipment should be unobtrusive and well integrated. The areas of application targeted in this thesis are activity monitoring and activity pattern discovery in apartments or small houses. The acquisition of information regarding the residents' activity is vital for the success of any assistive technology. In many areas of daily life, this is routine already. State-of-the-art sensing technology includes cameras, light barriers, RFID sensors, radio signal localization using transponders, and pressure sensitive Floors. Due to their operating principles, they have a big impact on home and living environments. Therefore, this thesis is dedicated to research for non-intrusive activity information acquisition technology, that has minimal impact on daily life. Two base technologies are taken into account in this thesis.

Nous abordons dans ces travaux l’identification non intrusive des charges des bâtiments résidentiels ainsi que la prédiction de leur état futur. L'originalité de ces travaux réside dans la méthode utilisée pour obtenir les résultats voulus, à savoir l'analyse statistique des données(algorithmes de classification). Celle-ci se base sur des hypothèses réalistes et restrictives sans pour autant avoir de limitation sur les modèles comportementaux des charges (variations de charges ou modèles) ni besoin de la connaissance des changements d'état des charges. Ainsi, nous sommes en mesure d’identifier et/ou de prédire l'état des charges consommatrices d'énergie (et potentiellement contrôlables) en se basant uniquement sur une phase d'entrainement réduite et des mesures de puissance active agrégée sur un pas de mesure de dix minutes, préservant donc la vie privée des habitants.Dans cette communication, après avoir décrit la méthodologie développée pour classifier les charges et leurs états, ainsi que les connaissances métier fournies aux algorithmes, nous comparons les résultats d’identification pour cinq algorithmes tirés de l'état de l'art et les utilisons comme support d'application à la prédiction. Les algorithmes utilisés se différencient par leur capacité à traiter des problèmes plus ou moins complexe (notamment la prise en compte de relations entre les charges) et se ne révèlent pas tous appropriés à tout type de charge dans le bâtiment résidentiel / Smart metering is one of the fundamental units of a smart grid, as many further applicationsdepend on the availability of fine-grained information of energy consumption and production.Demand response techniques can be substantially improved by processing smart meter data to extractrelevant knowledge of appliances within a residence. The thesis aims at finding generic solutions for thenon-intrusive load monitoring and future usage prediction of residential loads at a low sampling rate.Load monitoring refers to the dis-aggregation of individual loads from the total consumption at thesmart meter. Future usage prediction of appliances are important from the energy management point ofview. In this work, state of the art multi-label temporal classification techniques are implemented usingnovel set of features. Moreover, multi-label classifiers are able to take inter-appliance correlation intoaccount. The methods are validated using a dataset of residential loads in 100 houses monitored over aduration of 1-year.

Dans les maisons et bâtiments intelligents, il devient nécessaire de limiter l'intervention humaine sur le système énergétique, afin de fluctuer automatiquement l'énergie consommée par les appareils consommateurs. Pour cela, un système de mesure de la consommation électrique d'équipements est aussi nécessaire et peut être déployé de deux façons : intrusive ou non-intrusive. La première solution consiste à relever la consommation de chaque appareil, ce qui est inenvisageable à une grande échelle pour des raisons pratiques liées à l'entretien et aux coûts. Donc, la solution non-intrusive (NILM pour Non-Intrusive Load Monitoring), qui est capable d'identifier les différents appareils en se basant sur les signatures extraites d'une consommation globale, est plus prometteuse. Le problème le plus difficile des algorithmes NILM est comment discriminer les appareils qui ont la même caractéristique énergétique. Pour surmonter ce problème, dans cette thèse, nous proposons d'utiliser une information externe pour améliorer la performance des algorithmes existants. Les premières informations additionnelles proposées considèrent l'état précédent de chaque appareil comme la probabilité de transition d'état ou la distance de Hamming entre l'état courant et l'état précédent. Ces informations sont utilisées pour sélectionner l'ensemble le plus approprié des dispositifs actifs parmi toutes les combinaisons possibles. Nous résolvons ce problème de minimisation en norme l1 par un algorithme d'exploration exhaustive. Nous proposons également d'utiliser une autre information externe qui est la probabilité de fonctionnement de chaque appareil fournie par un réseau de capteurs sans fil (WSN pour Wireless Sensor Network) déployé dans le bâtiment. Ce système baptisé SmartSense, est différent de la solution intrusive car seul un sous-ensemble de tous les dispositifs est surveillé par les capteurs, ce qui rend le système moins intrusif. Trois approches sont appliquées dans le système SmartSense. La première approche applique une détection de changements de niveau sur le signal global de puissance consommé et les compare avec ceux existants pour identifier les dispositifs correspondants. La deuxième approche vise à résoudre le problème de minimisation en norme l1 avec les algorithmes heuristiques de composition Paréto-algébrique et de programmation dynamique. Les résultats de simulation montrent que la performance des algorithmes proposés augmente significativement avec la probabilité d'opération des dispositifs surveillés par le WSN. Comme il n'y a qu'un sous-ensemble de tous les appareils qui sont surveillés par les capteurs, ceux qui sont sélectionnés doivent satisfaire quelques critères tels qu'un taux d'utilisation élevé ou des confusions dans les signatures sélectionnées avec celles des autres. / In smart homes, human intervention in the energy system needs to be eliminated as much as possible and an energy management system is required to automatically fluctuate the power consumption of the electrical devices. To design such system, a load monitoring system is necessary to be deployed in two ways: intrusive or non-intrusive. The intrusive approach requires a high deployment cost and too much technical intervention in the power supply. Therefore, the Non-Intrusive Load Monitoring (NILM) approach, in which the operation of a device can be detected based on the features extracted from the aggregate power consumption, is more promising. The difficulty of any NILM algorithm is the ambiguity among the devices with the same power characteristics. To overcome this challenge, in this thesis, we propose to use an external information to improve the performance of the existing NILM algorithms. The first proposed additional features relate to the previous state of each device such as state transition probability or the Hamming distance between the current state and the previous state. They are used to select the most suitable set of operating devices among all possible combinations when solving the l1-norm minimization problem of NILM by a brute force algorithm. Besides, we also propose to use another external feature that is the operating probability of each device provided by an additional Wireless Sensor Network (WSN). Different from the intrusive load monitoring, in this so-called SmartSense system, only a subset of all devices is monitored by the sensors, which makes the system quite less intrusive. Two approaches are applied in the SmartSense system. The first approach applies an edge detector to detect the step-changes on the power signal and then compare with the existing library to identify the corresponding devices. Meanwhile, the second approach tries to solve the l1-norm minimization problem in NILM with a compositional Pareto-algebraic heuristic and dynamic programming algorithms. The simulation results show that the performance of the proposed algorithms is significantly improved with the operating probability of the monitored devices provided by the WSN. Because only part of the devices are monitored, the selected ones must satisfy some criteria including high using rate and more confusions on the selected patterns with the others.