Federated Learning-Based Resilient Control of Shipboard Power System

• Published in: IEEE Green Energy and Systems Conference pp. 1 - 7
• Main Authors: Bose, Anway, Amato, Joseph, Bai, Li
• Format: Conference Proceeding
• Language: English
• Published: IEEE 04.11.2024
• Subjects: Control systems; Federated learning; Federated Load Prediction; Federated Nested Regression; Federated XGBoost; Microgrids; Noise; Power system control; Power system reliability; Redundancy; Reliability engineering; Resilient Control; Robustness; Smart systems
• ISSN: 2640-0138
• DOI: 10.1109/GESS63533.2024.10784513

Shipboard power systems are analyzed and controlled using methods similar to those of terrestrial microgrids. One major difference between shipboard and terrestrial grids is that shipboard power system components are closely located to one another, relative to a more widely dispersed land-based system, and control systems leverage tightly coupled communication links as a benefit [1]. To mitigate the adverse effects of communication failures, Naval control systems are designed for redundancy and reliability, and various methods for isolating generation groups and power buses are implemented in the event of a communications disruption. However, isolated generator operations, even short-term, can limit the ship's capability to meet load demands and have significant operational impacts. In this paper, we propose a federated learning approach to implement decentralized load prediction and demonstrate that the results exhibit robustness in the presence of noise. The federated learning method for load prediction is a step toward designing more resilient Naval shipboard power system controls.