Enhancing wind power forecasting accuracy: a hybrid deep-learning approach amid curtailment scenarios

This study introduces a novel approach for wind power forecasting, addressing the unpredictability of wind energy production and the issue of power curtailment. We developed a hybrid model combining the strengths of Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and Transformer models, tailored to enhance the forecasting accuracy of wind power time series data. Through extensive experiments with data from two wind farms in Gansu Province, our model demonstrated superior performance over traditional models (ANN, SVR, RR, RF, LR) across four seasons, evidenced by lower RMSE and MAE values. Additionally, we proposed a systematic approach to manage power curtailment, effectively recovering curtailed power output values and ensuring data integrity for predictive modeling. Our findings not only contribute to the advancement of wind power forecasting methodologies but also highlight the potential for integrating advanced machine learning techniques to improve renewable energy management and grid stability.