A Multi-Node Energy Prediction Approach Combined with Optimum Prediction Interval for RF Powered WSNs.

Bikrant Koirala,Paul Keir,Keshav Dahal,Wenbing Chen

doi:10.3390/s19245551

Abstract

Energy prediction plays a vital role in designing an efficient power management system for any environmentally powered Wireless Sensor Networks (WSNs). Most of the Moving Average (MA)-based energy prediction methods depend on past energy readings of the concerned node to predict its future energy availability. However, in case of RF powered WSNs the harvesting history of the main node along with neighbouring nodes can also be used to develop a more robust prediction technique. In this paper, we propose a Multi-Node energy prediction method for Radio Frequency Energy Harvesting (RF-EH) WSNs, which predicts the future energy availability by taking into account harvesting history of all nodes surrounding the main node. We analyse the effective distance for prediction and also develop a mathematical model to compute the optimum value of prediction interval, which has a major effect in prediction accuracy and system design, considering energy neutrality. Results show that Multi-Node prediction is less sensitive to prediction interval while inheriting the advantages of MA techniques. Also, nodes located at a larger distance were utilized less for prediction, and as the prediction interval increased, the utilization of more distant nodes decreased. Furthermore, we also establish a linear relation between the prediction interval and the energy threshold limit.

Highlights

In recent times, there has been a growing interest in the development of Wireless Sensor Networks (WSNs) that are capable of harnessing energy from ambient sources, such as solar, wind, vibration, heat and electromagnetic waves [1,2]
To the best of our knowledge, no priorfor work has utilized energy readings from for RF-powered work has utilized energy neighbouring nodes in order to make energy predictions. no Theprior prediction approach takes into readings account from neighbouring nodes ordersurrounding to make energy predictions
The prediction takes the harvesting history of allinnodes the main node for which energyapproach prediction is tointo be account the harvesting history of all nodes surrounding the main node for which energy prediction made

Summary

Introduction

There has been a growing interest in the development of Wireless Sensor Networks (WSNs) that are capable of harnessing energy from ambient sources, such as solar, wind, vibration, heat and electromagnetic waves [1,2]. It was found that EWMA and the predictor developed at the Swiss Federal Institute of Technology of Zurich (ETHZ) occupied the least memory, while EWMA took the shortest time for prediction, with the average error in prediction being the smallest in the case of WCMA These works, only present prediction models based on a single node, and do not consider harvesting history of the neighbouring nodes while forecasting future energy availability for a particular node. In [16], the authors presented a model based on the prediction of future available energy for optimizing problems related to buffer sizes, timing, and rates by adapting the parameters of a solar-powered WSN. The literature described above does not provide sufficient insight for estimating an optimum prediction interval, which is crucial in developing prediction-based power management systems

Basic Models and Methods

Multi-Node Energy Prediction

Optimum Energy Prediction Interval

Results and Discussion

10. Neighbouring

Conclusions