Conserving Energy Through Neural Prediction of Sensed Data
The constraint of energy consumption is a serious problem in wireless sensor networks (WSNs). In this regard, many solutions for this problem have been proposed in recent years. In one line of research, scholars suggest data driven approaches to help conserve energy by reducing the amount of required communication in the network. This paper is an attempt in this area and proposes that sensors be powered on intermittently. A neural network will then simulate sensors’ data during their idle periods. The success of this method relies heavily on a high correlation between the points making a time series of sensed data. To demonstrate the effectiveness of the idea, we conduct a number of experiments. In doing so, we train a NAR network against various datasets of sensed humidity and temperature in different environments. By testing on actual data, it is shown that the predictions by the device greatly obviate the need for sensed data during sensors’ idle periods and save over 65 percent of energy.