| 초록 |
Efforts to apply renewable energy of wind and solar are steadily increasing to achieve the global climate targets. Because renewable energy supply is fluctuating over time, a co-production system composed of 4 units (renewable energy production site, energy storage system, power grid, and power-to-fuel process) is necessary to support flexible and continuous production of green fuel. However, the design and optimization of such systems coupled with renewable energy face challenges using conventional optimization methods, in which a constant operation policy is applied regardless of fluctuating renewable energy production or energy demand. This study aims to tackle this issue by adopting a reinforcement learning-based methodology to allow the optimizing policy to be changed in real-time. The new method allocates power flow between system components on an hourly basis with policies updated in real-time according to uncertain environments. The cost reduction by using the proposed method will be verified through a specific case study to which the Power-to-Ammonia (PtA) process is applied. |
