High-Severity N-x-k Contingency Ranking and Screening Based on Deep Learning and Heuristic Search

Contingency analysis is the important base and critical guarantee for security and reliability of hybrid AC/DC systems. This paper proposes a high-severity N-x-k contingency ranking and screening method for hybrid AC/DC systems. Firstly, the mechanism of successive commutation failures is analyzed and the minimum second voltage drop value of the commutation bus (V2nd c) after short-circuit fault is utilized as the index to quantify the severity of line outage contingencies. To quickly evaluate the impact of multi-line outage failures on DC systems, a deep learning-based assessment network of which the output is V2nd c and inputs are steady-state features related to network structure is built. Secondly, a two-stage heuristic search approach is proposed to screen and rank high-severity N-x-k line outage contingencies in hybrid AC/DC power systems. Stage 1 is to narrow search space and stage 2 is to evaluate the V2nd c and rank N-x-k contingencies based on the assessment network. A bus impedance matrix formation approach based on Woodbury formula is proposed, which is used in the calculation of input features and search index. Simulation results demonstrate that the proposed method can accelerate the screening and ranking of high-severity N-x-k line outage contingencies in hybrid AC/DC power systems.