A C-LMS Prediction Algorithm for Rechargeable Sensor Networks

• Published in: IEEE access Vol. 8; p. 1
• Main Authors: Ma, Dongchao, Zhang, Chenlei, Ma, Li
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Charging; Correlation coefficients; Energy harvesting; energy prediction; Model accuracy; network lifetime; Network topologies; Prediction models; Predictions; Rechargeable sensor network; Residual energy; Sensors; solar energy; Weather; Wireless networks; Wireless sensor networks
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2020.2986575

This paper focuses on the application environment of solar charging in Energy Harvesting Wireless Sensor Networks (EH-WSN), and studies how to effectively use energy prediction to extend the life of sensor networks. Considering the prediction algorithm of the standard Least Mean Square (LMS), the output power error is large when weather changes are fluctuating, and energy collection cannot be accurately predicted. This paper proposes a Correlation Least Mean Square (C-LMS) prediction model that introduces the correlation factor of weather changes. The algorithm has low complexity with a certain flexibility, which can solve it quickly and effectively improve the accuracy of short-term prediction. Experimental results show that the error rate of the C-LMS prediction algorithm is reduced by about 15% compared with the LMS model, and the prediction accuracy is significantly improved dealing with weather fluctuation. At the same time, based on the above lightweight prediction algorithm, the effects of predictive charging and residual energy on the rechargeable sensor network topology are reconsidered. Compared to a routing strategy that does not consider predictive charging, the optimized network lifetime has increased by nearly 31.7%.