A Composite Moving Average Algorithm for Predicting Energy in Solar Powered Wireless Sensor Nodes

Abstract

Recently, the operation of solar supplied wireless sensor nodes became dependent basically on implementing prediction algorithms for the expected harvested solar energy. These prediction algorithms contribute to saving energy. Thus, extending the lifetime of that sensors through utilizing the saved energy later. Among different prediction algorithms, the moving average based ones are the most used for their simplicity with limited resources in systems like sensor nodes. The basic moving average algorithm, Exponentially Weighted Moving Average (EWMA), is appropriate only to predict energies for symmetric days. Thus, it shows significant prediction error at times of sudden weather fluctuations. The other moving average algorithms, Weather Condition Moving Average (WCMA) and Pro-Energy show considerable errors specifically at sunset and sunrise times with a preference for the last one. This paper proposes a new moving average algorithm, named (EWWC), to eliminate the overall daily prediction error by combining two moving average algorithms (EWMA and WCMA). For two weeks, one in summer and another in winter, EWWC shows an average prediction error of (16.3%, 20.5%) while (EWMA) and (WCMA) show (21.2%, 25.1%) and (17.9%, 22.1%), respectively. This means that EWWC has an improvement of 1.6% in both weeks. Thus, it is recommended to be used for predicting energy in solar-powered sensor nodes.