Estimating Solar and Wind Power Production using Computer Vision Deep Learning Techniques on Weather Maps (2103.08727v2)

Abstract: Accurate renewable energy production forecasting has become a priority as the share of intermittent energy sources on the grid increases. Recent work has shown that convolutional deep learning models can successfully be applied to forecast weather maps. Building on this capability, we propose a ResNet-inspired model that estimates solar and wind power production based on weather maps. By capturing both spatial and temporal correlations using convolutional neural networks with stacked input frames, the model is designed to capture the complex dynamics governing these energy sources. A dataset that focuses on the state of California is constructed and made available as a secondary contribution of the work. We demonstrate that our novel model outperforms traditional deep learning techniques: it predicts an accurate power production profile that is smooth and includes high-frequency details.