Look-ahead decision making for renewable energy: A dynamic “predict and store” approach

This paper presents an integrative methodology for managing and stabilizing the output of a wind/solar farm using storage devices in cost effective real-time manner. We consider problem where renewable should decide amount energy charged into, or withdrawn from, battery given stochastic time-varying nature power output. Our features seamless integration non-myopic decision framework sequential non-parametric predictive model based on functional principal component analysis. A key feature our algorithm is that it quantifies costs over rolling horizon both predictions decisions are updated fly as new data acquired. technology tested California ISO dataset. The case study provides proof-of-concept highlights benefits ease implementation forward looking framework.