Cloud computing data centers play a critical role in storing, processing, and accessing vast amounts of digital information. They are essential infrastructures that support modern businesses, scientific research, communication networks, and various online services relying on reliable and efficient data management. However, their substantial energy consumption poses challenges in terms of operational costs and environmental sustainability. As a result, this paper investigates the utilization of distributed energy resources such as photovoltaics and wind turbines to reduce reliance on fossil fuels and achieve sustainability goals. Due to the intermittent and stochastic nature of these energy resources, traditional model-based methods are insufficient. Therefore, we propose a novel model-free Q-learning approach to address the challenges and ensure reliable and consistent power supply within data center operations. In the proposed structure, producers and consumers are modeled as autonomous agents within a multi-agent framework, making decisions to maximize their rewards. This paper addresses the hour-ahead energy scheduling problem using this approach, aiming to effectively manage the data center’s load demand and optimize distributed energy resources. By formulating the problem as a Markov Decision Process (MDP) and employing Q-learning, our study demonstrates increased energy producer profits and reduced data center costs. Finally, the proposed method is validated through simulation using real-world power datasets.