Falls in the elderly is a world health problem. Although many fall detection solutions were presented in literature, few of them are wrist-wearable devices, mainly due to typical processing and classification challenges to achieve accuracy greater than 95%. Considering the wrist as a more comfortable, discrete and acceptable place for an elderly wearable device, this work presents the development and evaluation of a wrist-worn fall detection solution. Different sensors (accelerometer, gyroscope and magnetometer), signals (acceleration, velocity and displacement) and direction components (vertical and non-vertical) were combined and a comprehensive set of threshold-based and machine learning methods were applied in order to define the best approach for fall detection. Data was acquired for fall and non-fall movements from twenty-two volunteers. For threshold-based methods, a maximum accuracy of 91.1% was achieved with 95.8% and 86.5% of sensitivity and specificity, respectively, using Madgwick’s decomposition. With the same movement decomposition and machine learning methods in the classification stage, an impressive accuracy of 99.0% was achieved, with 100% of sensitivity and 97.9% of specificity in our dataset. Prolonged tests with a volunteer wearing the fall detector also demonstrate the advantages of machine learning methods in terms of practical applications.

Photovoltaic (PV) energy use has been increasing recently, mainly due to new policies all over the world to reduce the application of fossil fuels. PV system efficiency is highly dependent on environmental variables, besides being affected by several kinds of faults, which can lead to a severe energy loss throughout the operation of the system. In this sense, we present a Monitoring System (MS) to measure the electrical and environmental variables to produce instantaneous and historical data, allowing to estimate parameters that ar related to the plant efficiency. Additionally, using the same MS, we propose a recursive linear model to detect faults in the system, while using irradiance and temperature on the PV panel as input signals and power as output. The accuracy of the fault detection for a 5 kW power plant used in the test is 93.09%, considering 16 days and around 143 hours of faults in different conditions. Once a fault is detected by this model, a machine-learning-based method classifies each fault in the following cases: short-circuit, open-circuit, partial shadowing, and degradation. Using the same days and faults applied in the detection module, the accuracy of the classification stage is 95.44% for an Artificial Neural Network (ANN) model. By combining detection and classification, the overall accuracy is 92.64%. Such a result represents an original contribution of this work, since other related works do not present the integration of a fault detection and classification approach with an embedded PV plant monitoring system, allowing for the online identification and classification of different PV faults, besides real-time and historical monitoring of electrical and environmental parameters of the plant.

In the subsequent decades, the increasing energy will demand renewable resources and intelligent solutions for managing consumption. In this sense, Non-Intrusive Load Monitoring (NILM) techniques detail consumption information for users, allowing better electric power management and avoiding energy losses. In high-frequency NILM methods, state-of-the-art approaches, mainly based on deep learning solutions, do not provide a complete NILM architecture, including all the required steps. To overcome this gap, this work presents an integrated method for detection, feature extraction, and classification of high-frequency NILM signals for the publicly available LIT-Dataset. In terms of detection, the results were above 90% for most cases, whilst the state-of-the-art methods were below 70% for eight loads. For classification, the final accuracies were comparable with other recent works (around 97%). We also include a multi-label procedure to avoid the disaggregation stage, indicating the loads connected at a given time, increasing the recognition of multiple loads. Finally, we present results in an embedded system, a subject also underexplored in the recent literature, demonstrating the proposal’s feasibility for real-time signal analysis and practical applications involving NILM.