Prediction of Harvested Energy for Wireless Sensor Node

Abstract

Energy harvesting wireless sensor nodes are interesting for the Internet of Things, since they can provide continuous operation by adapting workload not only to the current energy reserves, but to the amount of energy that can be harvested in the future also. We present a multistage day ahead hourly solar energy prediction algorithm. The predictor uses cloud cover and precipitation probability predictions from weather forecast obtained once per day for 24 hours in advance. To compensate for short-term weather changes until the next weather forecast data is obtained, forecast errors of humidity and atmospheric pressure are fed to the fuzzy logic filter. The filter adjusts predictions of cloud cover and precipitation probability, which are applied to the clear-sky radiation model in order to obtain prediction of solar energy. The prediction of solar energy is additionally corrected based on the energy prediction error in the preceding time slot. The results show that the proposed predictor outperforms the state-of-the-art predictors in terms of prediction error. Proposed predictor and state-of-the-art predictors were also evaluated using a simulated wireless sensor node with the simple energy management algorithm, where the proposed predictor was the most efficient at maintaining energy neutrality.