No abstract

Abstract-The Internet is smoothly migrating from an Internet of people towards an Internet of Things (IoT). By 2020, it is expected to have 50 billion things connected to the Internet. However, such a migration induces a strong level of complexity when handling interoperability between the heterogeneous Internet things, e.g., RFIDs (Radio Frequency Identification), mobile handheld devices, and wireless sensors. In this context, a couple of standards have been already set, e.g., IPv6, 6LoWPAN (IPv6 over Low power Wireless Personal Area Networks), and M2M (Machine to Machine communications). In this paper, we focus on the integration of wireless sensor networks into IoT, and shed further light on the subtleties of such integration. We present a real-world test bed deployment where wireless sensors are used to control electrical appliances in a smart building. Encountered problems are highlighted and suitable solutions are presented.

No abstract

Predicting energy consumption in Smart Buildings (SB), and scheduling it, is crucial for deploying Energy-efficient Management Systems. Most important, this constitutes a key aspect in the promising Smart Grids technology, whereby loads need to be predicted and scheduled in real-time to cope for the strongly coupled variance between energy demand and cost. Several approaches and models have been adopted for energy consumption prediction and scheduling. In this paper, we investigated available models and opted for machine learning. Namely, we use Artificial Neural Networks (ANN) along with Genetic Algorithms. We deployed our models in a real-world SB testbed. We used CompactRIO for ANN implementation. The proposed models are trained and validated using real-world data collected from a PV installation along with SB electrical appliances. Though our model exhibited a modest prediction accuracy, which is due to the small size of the data set, we strongly recommend our model as a blue-print for researchers willing to deploy real-world SB testbeds and investigate machine learning as a promising venue for energy consumption prediction and scheduling.

No abstract