Conformalized Prediction of Post-Fault Voltage Trajectories Using Pre-trained and Finetuned Attention-Driven Neural Operators

• DOI: 10.48550/arxiv.2410.24162
• Publication Date: 2024-10-31
• Authors: Amirhossein Mollaali; Gabriel Zufferey; Gonzalo E. Consta...; Christian Moya; Can Li; Guang Lin; Yue Meng
• Keywords: FOS: Computer and information sciences; Computer Science - Machine Learning; Machine Learning (cs.LG)

This paper proposes a new data-driven methodology for predicting intervals of post-fault voltage trajectories in power systems. We begin by introducing the Quantile Attention-Fourier Deep Operator Network (QAF-DeepONet), designed to capture complex dynamics and reliably estimate quantiles target trajectory without any distributional assumptions. The proposed operator regression model maps observed portion its unobserved trajectory. Our employs pre-training fine-tuning process address challenge limited data availability. To ensure privacy learning pre-trained model, we use merging via federated with from neighboring buses, enabling learn underlying such buses directly sharing their data. After pre-training, fine-tune bus, allowing it adapt unique operating conditions. Finally, integrate conformal prediction into fine-tuned coverage guarantees predicted intervals. evaluated performance using New England 39-bus test system considering detailed models frequency controllers. Two metrics, Prediction Interval Coverage Probability (PICP) Normalized Average Width (PINAW), are used numerically assess model's results show that approach offers practical reliable uncertainty quantification interval trajectories.