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.

MSc (Med), Dept of Haematology, School of Pathology, Faculty of Health Sciences, University of the Witwatersrand / BACKGROUND: Human Immunodeficiency Virus (HIV) is a global epidemic with growing numbers of people on highly active anti‐retroviral therapy (HAART) programmes. Effectiveness of treatment needs to be monitored to ensure the uncompromised well being of patients. This is currently done using both Viral Load (VL) and CD4 cell counts for HAART initiation and follow‐up. Although VL is the best predictor of disease progression it is often too expensive for monitoring patients in resource‐limited settings. There is thus a need for a cheaper, more accessible alternative to monitor long term patient response to therapy. METHODS: This study evaluated the use of a recently described flow cytometric assay of CD38 expression (previously developed at the Johannesburg Flow Cytometry Reference Laboratory) in a cohort of HIV+ patients failing 1st line therapy, who were subsequently enrolled onto 2nd line HAART. CD38 and CD8 were “piggy ‐backed” onto the PLG/CD4 protocol and mean fluorescence intensity (MFI) of the CD8/38 expression was monitored longitudinally. Patterns of CD38 expression were compared to 1st line treatment observations to establish equivalence in the predictive power of CD38 expression of fluctuation in viral load on 2nd line treatment patients. In addition, the effect of sample age on assay accuracy was tested before implementation of the CD38 assay at a secondary testing site. RESULTS: The patterns observed in the cohort of 2nd line therapy patients mirrored patterns previously seen in 1st line therapy with 55% of patients showing a continuous decline in CD38 MFI that mimicked changes in VL. The remaining 33% of patients had non‐specific increases in CD38 MFI without concurrent increases in VL and one patient showed irregular VL and CD38 MFI (non‐responder). The CD38 assay showed acceptable accuracy and reproducibility up to 48 hours after venesection (%CV<5%). Implementation at the secondary testing site was successful with 98% similarity (%CV<5%) compared to the reference laboratory. CONCLUSION: CD38 monitoring of 2nd line therapy patients showed comparable patterns to observations in 1st line therapy patients. The assay proved stable over time and easy to implement at another PLG/CD4 testing facility. As such, the CD38 assay offers a cost‐effective, reliable real time supplementary test to long‐term VL monitoring of HIV infected patients on the national ART programme.

In smart energy systems, load curve refers to the time series reported by smart meters, which indicate the energy consumption of customers over a certain period of time. The widespread use of load curve (data) in demand side management and demand response programs makes it one of the most important resources. To capture the load behavior or energy consumption patterns, load curve modeling is widely applied to help the utilities and residents make better plans and decisions. In this dissertation, with the help of load curve modeling, we focus on data-driven solutions to three load monitoring problems in different scenarios of smart energy systems, including residential power systems and datacenter power systems and covering the research fields of i) data cleansing, ii) energy disaggregation, and iii) fine-grained power monitoring. First, to improve the data quality for load curve modeling on the supply side, we challenge the regression-based approaches as an efficient way to load curve data cleansing and propose a new approach to analyzing and organizing load curve data. Our approach adopts a new view, termed portrait, on the load curve data by analyzing the inherent periodic patterns and re-organizing the data for ease of analysis. Furthermore, we introduce strategies to build virtual portrait datasets and demonstrate how this technique can be used for outlier detection in load curve. To identify the corrupted load curve data, we propose an appliance-driven approach that particularly takes advantage of information available on the demand side. It identifies corrupted data from the smart meter readings by solving a carefully-designed optimization problem. To solve the problem efficiently, we further develop a sequential local optimization algorithm that tackles the original NP-hard problem by solving an approximate problem in polynomial time. Second, to separate the aggregated energy consumption of a residential house into that of individual appliances, we propose a practical and universal energy disaggregation solution, only referring to the readily available information of appliances. Based on the sparsity of appliances' switching events, we first build a sparse switching event recovering (SSER) model. Then, by making use of the active epochs of switching events, we develop an efficient parallel local optimization algorithm to solve our model and obtain individual appliances' energy consumption. To explore the benefit of introducing low-cost energy meters for energy disaggregation, we propose a semi-intrusive appliance load monitoring (SIALM) approach for large-scale appliances situation. Instead of using only one meter, multiple meters are distributed in the power network to collect the aggregated load data from sub-groups of appliances. The proposed SSER model and parallel optimization algorithm are used for energy disaggregation within each sub-group of appliances. We further provide the sufficient conditions for unambiguous state recovery of multiple appliances, under which a minimum number of meters is obtained via a greedy clique-covering algorithm. Third, to achieve fine-grained power monitoring at server level in legacy datacenters, we present a zero-cost, purely software-based solution. With our solution, no power monitoring hardware is needed any more, leading to much reduced operating cost and hardware complexity. In detail, we establish power mapping functions (PMFs) between the states of servers and their power consumption, and infer the power consumption of each server with the aggregated power of the entire datacenter. We implement and evaluate our solution over a real-world datacenter with 326 servers. The results show that our solution can provide high precision power estimation at both the rack level and the server level. In specific, with PMFs including only two nonlinear terms, our power estimation i) at the rack level has mean relative error of 2.18%, and ii) at the server level has mean relative errors of 9.61% and 7.53% corresponding to the idle and peak power, respectively. / Graduate / 0984 / 0791 / 0800 / tangguo1999@gmail.com

Este trabalho apresenta uma ferramenta orientada ao objeto para o monitoramento de cargas em sistemas paralelos. O desenvolvimento desta ferramenta surgiu com o intuito de facilitar a programação paralela em sistemas distribuídos como NOWs, Networks of Workstations , e Grids computacionais, pois este tipo de programação é bem mais difícil do que a seqüencial e, por isso, desestimula novos programadores a desenvolver aplicações paralelas. Dentre as razões que tornam a programação paralela difícil destaca-se o balanceamento de cargas em que se quer maximizar a utilização dos recursos computacionais do sistema distribuído. Outro motivo para o programador de aplicações paralelas se preocupar com balanceamento de cargas é o desempenho, que é drasticamente afetado com o desequilíbrio de cargas do sistema. Com relação ao tempo em que as decisões de rebalanceamento de cargas são tomadas, os algoritmos de distribuição de cargas podem ser estáticos, realizados em tempo de compilação, ou dinâmicos, efetuados em tempo de execução. Embora o algoritmo estático não gere sobrecarga em tempo de execução na distribuição de carga, o dinâmico é a melhor escolha, pois se adapta bem em qualquer situação. Assim, o sistema de monitoramento de cargas surge como uma ferramenta de auxílio ao programador que deseje implementar algoritmos de balanceamento dinâmico de cargas nas suas aplicações paralelas, provendo informações de como os recursos computacionais do sistema distribuído estão sendo utilizados. / This work presents an object oriented tool for load monitoring in parallel systems. This tool was developed with intention to easy the parallel programming in distributed systems like NOWs (Networks of Workstations) and Computational Grids, because this type of programming is more difficult than the sequential and, therefore, it does not stimulate new programmers to develop parallel softwares. One of the most important reasons why parallel programming is difficult is the worry about load balancing where the purpose is to maximize the use of the computational resources of the distributed system. Another reason for the programmer of parallel softwares to worry about load balancing is the performance, which is drastically affected with the load imbalance of the system. With respect to the time where the decisions of load balancing are made, the load distribution algorithms can be static, done at compilation time, or dynamic, done at execution time. Although the static algorithm does not generate overhead at execution time, the dynamic one is a better choice, because it adapts well to any situation. Thus, the monitoring system appears as a tool to aid the programmer who desires to implement dynamic load balancing algorithms in his or her parallel softwares, providing information on how the computational resources of the distributed system are being used.

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%.

Este trabalho apresenta uma ferramenta orientada ao objeto para o monitoramento de cargas em sistemas paralelos. O desenvolvimento desta ferramenta surgiu com o intuito de facilitar a programação paralela em sistemas distribuídos como NOWs, Networks of Workstations , e Grids computacionais, pois este tipo de programação é bem mais difícil do que a seqüencial e, por isso, desestimula novos programadores a desenvolver aplicações paralelas. Dentre as razões que tornam a programação paralela difícil destaca-se o balanceamento de cargas em que se quer maximizar a utilização dos recursos computacionais do sistema distribuído. Outro motivo para o programador de aplicações paralelas se preocupar com balanceamento de cargas é o desempenho, que é drasticamente afetado com o desequilíbrio de cargas do sistema. Com relação ao tempo em que as decisões de rebalanceamento de cargas são tomadas, os algoritmos de distribuição de cargas podem ser estáticos, realizados em tempo de compilação, ou dinâmicos, efetuados em tempo de execução. Embora o algoritmo estático não gere sobrecarga em tempo de execução na distribuição de carga, o dinâmico é a melhor escolha, pois se adapta bem em qualquer situação. Assim, o sistema de monitoramento de cargas surge como uma ferramenta de auxílio ao programador que deseje implementar algoritmos de balanceamento dinâmico de cargas nas suas aplicações paralelas, provendo informações de como os recursos computacionais do sistema distribuído estão sendo utilizados. / This work presents an object oriented tool for load monitoring in parallel systems. This tool was developed with intention to easy the parallel programming in distributed systems like NOWs (Networks of Workstations) and Computational Grids, because this type of programming is more difficult than the sequential and, therefore, it does not stimulate new programmers to develop parallel softwares. One of the most important reasons why parallel programming is difficult is the worry about load balancing where the purpose is to maximize the use of the computational resources of the distributed system. Another reason for the programmer of parallel softwares to worry about load balancing is the performance, which is drastically affected with the load imbalance of the system. With respect to the time where the decisions of load balancing are made, the load distribution algorithms can be static, done at compilation time, or dynamic, done at execution time. Although the static algorithm does not generate overhead at execution time, the dynamic one is a better choice, because it adapts well to any situation. Thus, the monitoring system appears as a tool to aid the programmer who desires to implement dynamic load balancing algorithms in his or her parallel softwares, providing information on how the computational resources of the distributed system are being used.

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.

This thesis presents the development of techniques which enable appliance recognition in an Advanced Electricity Meter (AEM) to aid individuals reduce their domestic electricity consumption. The key aspect is to provide immediate and disaggregated information, down to appliance level, from a single point of measurement. Three sets of features including the short term time domain, time dependent finite state machine behaviour and time of day are identified by monitoring step changes in the power consumption of the home. Associated with each feature set is a membership which depicts the amount to which that feature set is representative of a particular appliance. These memberships are combined in a novel framework to effectively identify individual appliance state changes and hence appliance energy consumption. An innovative mechanism is developed for generating short term time domain memberships. Hierarchical and nearest neighbour clustering is used to train the AEM by generating appliance prototypes which contain an indication of typical parameters. From these prototypes probabilistic fuzzy memberships and possibilistic fuzzy typicalities are calculated for new data points which correspond to appliance state changes. These values are combined in a weighted geometric mean to produce novel memberships which are determined to be appropriate for the domestic model. A voltage independent feature space in the short term time domain is developed based on a model of the appliance’s electrical interface. The components within that interface are calculated and these, along with an indication of the appropriate model, form a novel feature set which is used to represent appliances. The techniques developed are verified with real data and are 99.8% accurate in a laboratory based classification in the short term time domain. The work presented in this thesis demonstrates the ability of the AEM to accurately track the energy consumption of individual appliances.

Mechanical durability of the structures should be continuously monitored during their operation. Structural health monitoring (SHM) techniques are typically used for gathering the information which can be used for evaluating the current condition of a structure regarding the existence, location, and severity of the damage. Damage can occur in a structure after long-term operating under service loads or due to incidents. By detection of these defects at the early stages of their growth and nucleation, it would be possible to not only improve the safety of the structure but also reduce the operating costs. The main goal of this dissertation is to develop a reliable and cost-effective SHM system for inspection of composite and metallic structures. The Surface Response to Excitation (SuRE) method is one of the SHM approaches that was developed at the FIU mechatronics lab as an alternative for the electromechanical impedance method to reduce the cost and size of the equipment. In this study, firstly, the performance of the SuRE method was evaluated when the conventional piezoelectric elements and scanning laser vibrometer were used as the contact and non-contact sensors, respectively, for monitoring the presence of loads on the surface. Then, the application of the SuRE method for the characterization vii of the milling operation for identical aluminum plates was investigated. Also, in order to eliminate the need for a priori knowledge of the characteristics of the structure, some advanced signal processing techniques were introduced. In the next step, the heterodyne method was proposed, as a nonlinear baseline free, SHM approach for identification of the debonded region and evaluation of the strength of composite bonds. Finally, the experimental results for both methods were validated via a finite element software. The experimental results for both SuRE and heterodyning method showed that these methods can be considered as promising linear and nonlinear SHM approaches for monitoring the health of composite and metallic structures. In addition, by validating the experimental results using FEM, the path for further improvement of these methods in future researches was paved.

Additional electricity meters behind the grid access point can improve understanding of energy consumption patterns and thus, adjust consumption behavior. For this study, smart meters were installed in three hostels, out of which two are analyzed further in this paper. Starting from an onsite inspection, all appliances were assigned to reasonable groups for sub-metering. Based on data for the year 2021, the sites are characterized according to the sub-metering concept. In addition, load profiles for type-days are derived, which allows to establish a baseload during COVID lockdown and compare it to consumption patterns for normal occupation. In the prescriptive part, the demand profiles are analyzed regarding their economic potential for load shifting. Consumption data for one week with normal occupation is used as input for techno-economic modeling. The mixed-integer model minimizes electricity purchasing costs for different scenarios including dynamic tariffs and onsite generation from photovoltaics.