In the era of exponential data growth, data centers play a critical role in providing reliable and efficient computing resources. However, the energy consumption of these facilities has become a significant concern, driving the need for innovative solutions to enhance energy efficiency. This paper presents a novel approach to minimizing virtual machine (VM) migrations in data centers, thereby reducing energy consumption. Our model leverages a Recurrent Neural Network (RNN) optimized with a Grey Wolf Optimizer (GWO) to predict energy consumption and preemptively manage workloads. By incorporating this RNN, our model effectively anticipates the energy consumption of future, leading to make the best decision to migrate the VM or not. Furthermore, energy consumption is monitored in real-time to identify overloading and underloading of VMs. The reinforcement GWO-RNN framework adapts dynamically to changing workloads, ensuring optimal resource allocation and reducing the frequency of energy-intensive VM migrations. Experimental results demonstrate a significant reduction in energy consumption and improved efficiency in data center operations, showcasing the potential of this approach for sustainable computing.