Reducing energy consumption and increasing reliability in wireless sensor networks using artificial intelligence

Energy consumption in general is one of biggest challenges when it comes to wireless sensor networks (WSNs). Since the biggest amount of energy is used for communication, the most logical way to reduce the energy consumption is to reduce the number of packets transmitted between sensor and sink node. To address this issue, data reduction methods, which are predicting the measured values both at source and sink node, have been developed. Consequently, transmitting measurements is only required if the sensed value differs from the predicted one by a given threshold. The choice of data reduction strategy depends on the type of sensed phenomena. Especially when trying to predict data of nonlinear systems with a high entropy like electric engine or gearbox vibration, it becomes very difficult to build a model that describes and forecasts those values. In this paper, we use a Self-Exciting Threshold Autoregressive (SETAR) model as a Time Series Forecasting method in combination with nonlinear systems to overcome that problem. By applying this algorithm on real-world data using panStamp technology, we achieve a maximum energy saving of up to 73%.

network, reducing energy consumption, increasing reliability.