SIPS: Solar Irradiance Prediction System

Abstract

There is high interest in up-scaling capacities of renewable energy sources such as wind and solar. However, variability and uncertainty in power output is a major concern and forecasting is, therefore, a top priority. Advancements in forecasting can potentially limit the impact of fluctuations in solar power generation, specifically in cloudy days when the variability and dynamics are the largest. We propose SIPS, Solar Irradiance Prediction System, a novel sensing infrastructure using wireless sensor networks (WSNs) to enable sensing of solar irradiance for solar power generation forecasting. In this paper, we report the findings of a deployment of a hierarchical WSN system consisting of 19 TelosB nodes equipped with solar irradiance sensors, and 5 MicaZ nodes equipped with GPS boards, deployed in the vicinity of a 1 MW solar array. We evaluate different irradiance sensor types and the performance of different novel prediction methods using SIPS' data and show that the spatial-temporal cross-correlations between sensor node readings and solar array output power exists and can be exploited to improve prediction accuracy. Using this data for short-term solar forecasting for cloudy days with very high dynamics in solar output power generation - the worst case scenario for prediction-, we get an average of 97.24% accuracy in our prediction for short time horizon forecasting and 240% reduction of predicted normalized root mean square error (NRMSE) compared to state-of-the-art methods that do not use SIPS data.