Improving energy usage in energy harvesting wireless sensor nodes using weather forecast

Abstract

Abstract -Battery powered wireless sensor nodes are used in many applications. They can be placed in remote locations and can operate for a long period of time without the need for maintenance. However, finite battery lifetime is one of the main limitations of such devices. Energy harvesting can be used to recharge batteries from environmental energy sources. This enables continuous work when energy neutrality is satisfied. For optimal use of available energy, when energy harvesting is used, the optimization goal switches from energy management to workload maximization while maintaining energy neutrality. In order to achieve energy neutrality, prediction of energy that can be harvested in the future is needed. This prediction can be based on previous measured data. However, this approach can be unreliable when weather conditions change during the day or between days. To improve prediction precision, weather forecast can be used. This information has been used to predict energy that can be harvested in the future but only for the next few hours. We present a twolevel predictor that uses cloud cover information from hourly weather forecast for next 24 hour period to predict energy that can be harvested in the same time interval. Proposed predictor achieves a 26% less prediction mean absolute percentage error, a 15% less mean absolute deviation percent error and allows an 8% better performance of simulated wireless sensor node compared to Exponentially Weighted Moving Average (EWMA) predictor.

Prediction of Harvested Energy for Wireless Sensor Node

“…Proposed predictor was compared to Pro-Energy [11], IPro-Energy [12], and our previous work [19] in terms of prediction error, as well on achieving energy neutrality.…”

“…In our previous paper [19], we have used Photovoltaic Geographical Information System (PVGIS) [20] to obtain cloudless sky solar irradiance data. PVGIS provides monthly, daily, and hourly solar irradiance data.…”

“…This step enables correction of forecast data for weather changes during the night. Proposed predictor has been compared to Pro-Energy and IPro-Energy, state-of-the-art PPF predictors, and our previous work [19] in terms of prediction error, and it gives better prediction results. Additionally, all predictors have been tested on a simulated wireless sensor node with a simple energy management algorithm in order to evaluate predictors' ability to maintain energy neutrality.…”

“…Instead of limiting  as in [19], we use the cloudless sky solar energy to limit the predicted energy after correction. This way, even if overestimation occurs, it is bound.…”

“…As stated previously, the cloud cover is used as a parameter, which modulates clear-sky direct irradiance. Compared to our previous work [19], we have also included precipitation probability, since it has been shown that it can affect solar radiation reaching Earth's surface [21]. We used precipitation probability data to modulate the clear-sky diffuse irradiance .…”