A new SADIT-MAR lecture titled “Intelligent Energy Management for Hybrid Propulsion Systems” by Dr. Peng Wu has been published, offering a comprehensive and up-to-date perspective on the decarbonization of maritime propulsion through hybrid technologies. In this lecture, the motivation, system-level challenges, and advanced control approaches for hybrid propulsion systems are systematically addressed. State-of-the-art methods based on optimization and deep reinforcement learning are introduced, and their applicability is demonstrated through realistic marine case studies. The lecture brings together theory, modeling, and real-world implementation within a unified and generic energy management framework.
Key topics covered in the lecture include
- The motivation for decarbonizing the maritime industry is introduced, with greenhouse gas, NOx, and SOx emissions from global and coastal shipping highlighted, along with the pressure to achieve significant emission reductions by 2050.
- Marine propulsion using batteries and fuel cells is reviewed, including current battery-powered and hybrid ship applications, relevant fuel cell technologies for marine use, and key limitations inalong with operational concepts such as plug-in operation, energy storage, and the supply of auxiliary and propulsion loads energy density, degradation, and system integration.
- Hybrid propulsion systems combining fuel cells, batteries, and shore power are presented, along with operational concepts such as plug-in operation, energy storage, and the supply of auxiliary and propulsion loads.
- The formulation of the energy management problem as a Markov decision process is explained, showing how a generic, multi-objective energy management framework for hybrid propulsion systems is established.
- The use of reinforcement learning and deep reinforcement learning for optimal power distribution between fuel cells and batteries is demonstrated, and performance close to the offline optimum under uncertainty is highlighted.
- A generic energy management framework is demonstrated through application to a real ship, and its potential for extension beyond marine applications to other sectors is emphasized.
