Combinatorics

Physics

Power (physics)

Interval (graph theory)

Mathematics

Environmental science

Term (time)

Quantum mechanics

Energy

With the increasing penetration of wind energy, accurate wind power prediction is essential for efficient utilization, equipment protection, and stable grid-connection of wind energy. Prediction interval (PI) provides a practical way to quantify the potential uncertainties of wind power. High coverage probability and narrow width are expected when constructing PI, however, the conflicts between these two objectives challenge the prediction accuracy. Therefore, a comprehensive PI quality evaluation strategy is presented in this paper to make a trade-off between coverage probability and interval width. It enables the prediction model to be trained by algorithms with high computational efficiency. Based on this evaluation strategy, a short-term wind power interval prediction method, which combines attention mechanism-based gated recurrent unit (Att-GRU), variational mode decomposition (VMD), and rolling fuzzy granulation (RFG), is proposed to construct high-quality PI and accurately capture the uncertainty of wind power. The validity of the proposed method is verified by comparative case studies on different datasets. Compared with other benchmark methods, the interval width is reduced by at least 22.66%, and the coverage probability and computation time have been promoted by at least 3.03% and 4.31% respectively. • Two modified PI quality evaluation indices PINCP and PINCW are proposed. • An improved cost function SCWC is defined to reduce the training complexity. • A VMD-RFG feature extraction method is designed to construct training labels. • An interval prediction method is proposed to capture the wind power uncertainty.

Short-term wind power interval prediction method using VMD-RFG and Att-GRU