Short-Term Wind Power Prediction Based on DTW-BiLSTM Modeling

Wind power forecasting plays a critical role in wind farm planning, operations, maintenance, and market trading. It improves the efficiency of wind energy production, minimizes system uncertainties, and provides a solid foundation for evaluating the economic viability of wind power and integrating renewable energy into the grid. However, the complex, non-linear nature of wind power data makes it challenging to capture the relationships within time series.To address these challenges, this study introduces a DTW-BiLSTM approach for short-term wind power forecasting. This method first captures the non-linear dependencies in the time series data through clustering, then extracts temporal features for each cluster. To enhance the accuracy and efficiency of forecasting, the dataset is normalized, and similar days are grouped using a sliding window technique. A BiLSTM model is then employed to fully capture the temporal correlations within the wind power data. To evaluate the effectiveness of this approach, the model was compared against LSTM, TCN, and CNN using performance metrics such as MAE, MSE, and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>. The results show that the proposed method outperforms the alternatives, offering more accurate and reliable predictions for wind power forecasting.