Spatio-Temporal Deep Learning-Assisted Multi-Period AC Optimal Power Flow

Abstract

The increasing penetration of renewable energy resources has amplified variability and uncertainty in power systems, reducing the effectiveness of conventional single-period Optimal Power Flow (OPF) strategies. Multi-period AC OPF (MP-ACOPF) offers a more comprehensive framework by incorporating inter-temporal constraints and resource flexibility, but its high computational complexity and strong temporal coupling make large-scale applications challenging, often causing scalability issues and convergence difficulties in conventional solvers. To overcome these limitations, we propose a deep learning approach that integrates Graph Attention Networks (GAT) and Temporal Convolutional Networks (TCN) to jointly capture spatial and temporal dependencies in MP-ACOPF. The proposed model is trained and evaluated on large-scale systems, including 500-bus and 1354-bus networks, under both 8-period and 24-period scenarios, demonstrating robust scalability and consistently high prediction accuracy. Using the model’s predicted outputs as an initial solution for conventional OPF solvers improves their convergence performance, demonstrating its effectiveness as an auxiliary tool for solving complex MP-ACOPF problems.

Keywords: Multi-period optimal power flow, AC optimal power flow, Deep Learning, Graph neural networks