Joint Probabilistic Forecasting of Wind and Solar Power Exploiting Spatiotemporal Complementarity
Fahong Zhang, Zhiyuan Leng, Lu Chen, Yongchuan Zhang
Abstract
Reliable and precise joint probabilistic forecasting of wind and solar power is crucial for optimizing renewable energy utilization and maintaining the safety and stability of modern power systems. This paper presents an innovative joint probabilistic forecasting model designed to address probabilistic spatiotemporal power output forecasting challenges. Leveraging a multi-network deep learning framework, the model integrated the temporal convolutional network for temporal feature extraction, the convolutional neural network for spatial feature analysis, and the attention mechanism for spatiotemporal focus enhancement, thereby capturing the spatiotemporal complementarity of wind and solar power. It also incorporated a quantile regression-based uncertainty quantification technique, contributing to reliable probabilistic predictions. A wind farm and two solar farms in China were used as a case study. Comparison results between the proposed model and ten established models demonstrated its superior performance in both reliable deterministic and probabilistic predictions, offering valuable insights for sustainable and resilient energy system operation.