부산대학교 도서관

Optimizing energy costs in Smart Grids is a critical task to meet increasing energy demands and efficiently integrate renewable sources into the grid. This paper explores the application of reinforcement learning, specifically Deep Reinforcement Learning (DRL), to enhance energy efficiency and reduce energy costs. Utilizing Deep Q-Learning algorithms, we demonstrate how this framework offers an effective model able to learn the optimal policy for selecting the best actions in a non-stationary environment. The results indicate that the Deep Q-Learning algorithm effectively decreases energy costs by optimally managing existing energy sources and minimizing losses. The performance of the proposed model is evaluated and compared with simpler baseline algorithms. The comparison in terms of accuracy and computation time reveals a significant advantage of using the Deep Q-Learning algorithm. Additionally, we discuss insights into the benefits and challenges of applying reinforcement learning to smart grid problems, providing guidance for future research.