Optimization of power and energy management for fuel cell-fed hybrid electric system in marine applications

• Main Author: Chen, Wenjie
• Other Authors: Tai Kang
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Hybrid power system; Optimization of power and energy management; Renewable energy; Energy storage; Fuel cell; Marine shipboard power system
• Online Access: https://hdl.handle.net/10356/174631
• Institution: Nanyang Technological University

Environmental sustainability has become a significant policy concern in global maritime transport in recent years. To achieve high energy efficiency and low emissions, the all-electric ship (AES) integrated with an energy storage system (ESS) is believed to be one of the most promising technologies for complying with environmental regulations. The traditional rule-based power management system (PMS) is not able to handle the complexity of this new shipboard power network configuration or even achieve optimal control. Advanced power management control is required to confront the new challenges of the AES hybrid power grid. The main objective of this PhD research is to develop an improved PMS strategy to achieve optimal operation and minimize total cost of ownership (TCO) and operation of marine vessels, considering fuel efficiency, emission limits and the lifetime of power devices. In this study, a system-level fuel cell-fed shipboard power plant with DC distribution is developed with MATLAB/Simulink platform. A hardware-in-the-loop (HIL) has been set up to replicate the real-time system behaviour. Both the mathematical and HIL models are validated against the full-scale shipboard power system. In addition, a unique optimization problem formulation for shipboard power management has been proposed and demonstrated for the first time, minimizing an objective function incorporating not just fuel consumption but also lifecycle cost of the power devices and penalty cost of emissions, all expressed in monetary terms. An improved supervisory real-time optimization-based PMS is proposed with two different approaches: model predictive control (MPC)-based and reinforcement learning (RL)-based power management strategies. An adaptive MPC (AMPC) with a novel hierarchical architecture that includes a mode selection component is designed to optimize the power allocation between different power sources of the shipboard power plant to achieve cost-effective multi-objective control. It is also a robust and reliable control that can handle load fluctuations and disturbances to improve system stability. On the other hand, a novel RL-based PMS control is also explored to apply the model-free, off-policy deep deterministic policy gradient (DDPG) algorithm to support continuous action space control for the first time. The feasibility and control performance of the proposed optimization-based PMS is validated against the HIL plant with a typical tugboat’s operating profiles as a case study. The advantages and cost analysis of the proposed strategies are compared against a traditional rule-based control system and a theoretical operation as the baselines. Compared with the traditional rule-based PMS, the proposed AMPC and RL approaches can achieve significant savings of up to 12.19% and 12.01% of TCO, respectively, and zero power device replacement throughout the ten years of long-term vessel operation under zero emission operation mode.