Document Type : Research Articles
Authors
1 Department of Electrical and Electronic Eng, University of Sistan and Baluchestan
2 Department Of Marine Engineering, Chabahar Maritime University, Chabahar, Iran
Abstract
Generator faults can partition shipboard microgrids into isolated zones, making energy storage system (ESS) placement as consequential as capacity for maintaining critical loads. This study jointly optimizes ESS placement and sizing in an eight-bus zonal shipboard microgrid, balancing cost, weight, and critical-load outage risk. Resilience is quantified using the conditional value-at-risk (CVaR) of a served-energy index under uncertain wind generation and demand. A conditional generative adversarial network produces fault-window scenarios with variability calibrated to historical data, while a neural-network surrogate replaces scenario-specific load-shedding computations within NSGA-II. The selected knee-point design locates storage in the critical forward zone, increasing worst-case resilience from 0.794 without ESS to 1.000. Because this design saturates the resilience objective at the full weight budget, statistical significance is evaluated using a representative, non-saturating sub-budget design at the same location. Relative to its own no-ESS baseline, this design improves worst-case resilience by 0.076 (95% CI [0.075, 0.078]); a paired Wilcoxon signed-rank test on matched per-scenario resilience indices yields p < 10⁻³⁴. The surrogate achieves a test R² of 0.871, with a maximum absolute error of 0.095 on a held-out subset of 50 designs, and accelerates objective evaluation approximately 1472-fold. Pareto-front stability across 20 random seeds is reflected in a hypervolume of 1.196 ± 0.002. These results demonstrate an efficient framework for storage planning that accounts for fault-induced isolation and uncertain operating conditions.
Keywords
Main Subjects
)