Predição de dados de sensoriamento visando eficiência energética através da redução de transmissão de dados em redes de sensores sem fio

Abstract

Wireless sensor networks (RSSF) are already of interest to researchers and are becoming increasingly integrated with home, commercial, industrial, etc. systems. However, due to its own characteristics, the problem of power restriction and management in the grid itself is one of the main points to be equalized. In this context, there are several approaches to provide energy efficiency, and in view of this, this work aims to compare prediction models of sensing data in wireless sensor networks in order to save energy in data transmission. ARIMA, SVM, MLP/ANN and RNN/LSTM models were used in a temperature data collection application and evaluated for the energy savings provided. The measurements were performed by real devices and it was possible to observe the performance of the models for a sensing data set in an office environment. For this work, the SVM model presented better performance compared to the others in terms of energy efficiency.