[M-ZZ40] Renewable energy and earth science

[MZZ40-03] Quantification of Uncertainty Inherent in Photovoltaic Output Prediction

*Daisuke Nohara¹, Yuki Kanno¹ (1.Central Research Institute of Electric Power Industry)

Keywords:photovoltaic, prediction, uncertainty

Renewable energy such as photovoltaic (PV) and wind power have a character of volatility due in part to the time evolution of weather systems. Prediction for the PV power outputs is one of the most cost-effective and easily implemented tools. Despite the recent increase in the accuracy of numerical weather prediction models, there is a limitation to reducing the prediction error. In addition to the improvement of the models, the use of statistical methods such as machine learning is accelerating. In this study, we summarize the causes of the prediction error of the PV and quantification of the uncertainty inherent in the prediction. The uncertainty caused by the chaotic behavior of the atmosphere increases with prediction time, but it is difficult to reduce the uncertainty. On the other hand, systematic uncertainties are caused by imperfections in the models, PV power conversion, and other factors. The uncertainties can be reduced by statistical methods. The ratio of uncertainties in the chaotic behavior and systematic is about 1 to 2 (2 to 3) for the 0-6 (24-30) hours ahead prediction. Systematic uncertainties can be reduced using statistical methods, suggesting that there is still room to reduce the spread of probabilistic predictions and prediction errors.