Deep learning-based multi-output quantile forecasting of PV generation
This work addresses the problem of accurate and efficient PV generation forecasting for energy grid operators, but it is incremental as it applies existing deep learning methods to a specific domain with tailored improvements.
The paper tackles probabilistic forecasting of photovoltaic (PV) generation by developing a deep learning-based encoder-decoder model that computes intraday multi-output quantile forecasts, using weather data as inputs and quantile regression for training. Results show improved forecast quality, as assessed by continuous ranked probability and interval scores, and computational efficiency for integration into intraday decision-making tools.
This paper develops probabilistic PV forecasters by taking advantage of recent breakthroughs in deep learning. It tailored forecasting tool, named encoder-decoder, is implemented to compute intraday multi-output PV quantiles forecasts to efficiently capture the time correlation. The models are trained using quantile regression, a non-parametric approach that assumes no prior knowledge of the probabilistic forecasting distribution. The case study is composed of PV production monitored on-site at the University of Liège (ULiège), Belgium. The weather forecasts from the regional climate model provided by the Laboratory of Climatology are used as inputs of the deep learning models. The forecast quality is quantitatively assessed by the continuous ranked probability and interval scores. The results indicate this architecture improves the forecast quality and is computationally efficient to be incorporated in an intraday decision-making tool for robust optimization.