PRICE-AWARE SHORT-TERM WIND POWER FORECASTING FOR INTRADAY MARKET TRADING

抄録

Short-term wind power forecasting is of vital importance for intraday power trading by assisting wind farm operators in making appropriate power selling plans and reducing the foreseen imbalances. However, the eventual imbalance penalties are affected not only by the discrepancy between the planned and actual values of wind generation but also by the market price. Specifically, a not-very-accurate forecast may be acceptable when the power price is low and the imbalance penalty price is correspondingly low, but a rather accurate forecast will be highly demanded otherwise. On the other hand, most of the related studies have focused mainly only on improving average prediction accuracy. In this viewpoint, a novel price-aware wind power forecasting method based on a long short-term memory neural network (LSTM-NN) with customized loss function for training is proposed, aiming to contribute to reducing the imbalance penalties that depend on the electricity market price.