A multi-task transient stability assessment method adapted to topology changes using multi-graph sample and aggregate-attention network

Abstract

Transient stability assessment (TSA) plays a pivotal role in guiding power grid risk control strategies. However, it faces challenges when dealing with complex multi-graph inputs generated by pre-fault, fault occurrence, and post-fault states. Meanwhile, most previous research studies neglected the assessment of the transient stability level. To address this, we propose a multi-task transient stability assessment (MTTSA) approach. In MTTSA, we introduce a multi-graph sample and aggregate-attention network (GraphSAGE-A) designed to capture stability features even amidst topology changes. A multi-head attention mechanism and local normalization method are adopted for a better extraction of the global and contextual information. Additionally, we introduce a quantified transient stability risk index considering the transient stability boundary and incorporate a multi-task dense structure to enhance MTTSA’s performance. Empirical tests, under changing operating conditions, conducted on the IEEE 39-bus system showcase a significant performance improvement with the proposed MTTSA method.