Very Short-term Prediction for Wind Power Based on BiLSTM-Attention

Abstract

With the increasing permeability of wind power (WP) in power systems, WP randomicity bring enormous difficulty to the dispatching departments of power grid. Exact description of WP is key to reduce the threat of uncertainty to power systems. First, Pearson correlation analysis of WP historical data and numerical weather prediction (NWP) is performed to preprocess and establish datasets. Then, the attention mechanism is integrated to improve bi-directional long short-time memory (BiLSTM) and establish very short-term prediction approach for WP based on improved BiLSTM. Finally, different evaluation indexes are used to test the practicability of the BiLSTM-Attention model in practical engineering application.