Correlation-based feature selection for resilience analysis of MVDC shipboard power system

Abstract

The future all-electric ship (AES) requires a resilient shipboard power system (SPS) that can withstand extreme events without substantial damage and maintaining critical capabilities. For superior capability to sustain disruptions, numerical measures of resilience provide assistance to engineers in making planning and operational decisions. This paper presents a novel method for analyzing the resilience characteristics of an SPS using correlation-based feature selection (CFS) to identify the system attributes that are the best predictors of performance during contingencies. Such attributes are typically chosen in the design stage and are not subject to modification during operation. The selected features are adjusted and their ramifications on the SPS performance are evaluated under the same contingencies. Results show quantitatively how a change of relevant attributes can improve or degrade the system performance in terms of unserved load and minimum recovery time.