Short-Term Wind Power Forecasting Method Based on NWP Wave Division and IMTCN-Seq2Seq

High-precision wind power prediction is crucial for the grid integration of wind power. To improve forecasting accuracy, this paper proposes a short-term wind power forecasting method based on Numerical Weather Prediction (NWP) wave division and an improved Multi-Scale Temporal Convolution Network-Seq2Seq (IMTCN-Seq2Seq) model. The method first applies wavelet transform to filter meteorological sequences most relevant to power, extracting fluctuation trends. Then, it divides the meteorological and power waves based on the extrema points of the trend sequences. Next, the improved multi-scale temporal convolution network extracts hidden features from the divided meteorological waves. Finally, the Seq2Seq model with attention mechanism generates the wind power forecast. Experimental results show that, compared to baseline models, this method significantly improves forecasting accuracy.