A Method for Wind Power Generation Interval Prediction in Cold Wave Conditions Based on Generative Adversarial Networks and Convolutional Neural Networks

In recent years, the global shortage of fossil fuels and worsening environmental issues have led countries worldwide to opt for renewable energy sources like wind power. Wind energy is significantly influenced by meteorological conditions, particularly during cold waves, where power output can exhibit irregular fluctuations, potentially causing disruptions to the power system and threatening its stable operation. Consequently, forecasting wind power generation intervals during cold waves is of paramount importance for ensuring grid safety. To address the scarcity of cold weather data samples, this paper employs Generative Adversarial Networks (GAN) to augment the limited dataset. Following data expansion, an interval prediction model for wind power generation under cold wave conditions is developed. The effectiveness of the method is validated by comparing it with the Extreme Learning Machine (ELM) model, and the proposed GAN-CNN method has better prediction performance than GAN-ELM. The key innovation of this paper lies in addressing the lack of cold wave data by incorporating a technique for small sample expansion, thereby providing a solid data foundation for wind power forecasting under such conditions.