Federated Learning (FL) in Vehicular Edge Computing (VEC) plays a key role in the development of Intelligent Transportation Systems (ITS), enhancing real-time decision-making and preserving data privacy in connected vehicle networks. However, the dynamic and resource constrained nature of VEC, coupled with non-IID data distributions, presents significant challenges to the effective deployment of FL. Current solutions often use static or greedy scheduling methods that lack adaptability, resulting in suboptimal learning performance, inefficient energy utilization, and increased latency. This study proposes a novel Data Aware Deep Reinforcement Learning (DRL) algorithm, namely DA-FeDRL, which utilizes an improved Twin Delayed Deep Deterministic Policy Gradient (TD3) method to optimize device scheduling while achieving optimal resource allocation by balancing data diversity, energy efficiency, and latency optimization. A key contribution of this work is the development of a reward function that jointly minimizes energy consumption and time delay while integrating Data Diversity metrics to enhance learning performance. By expanding the state representation to include crucial factors like device energy, data diversity, and network conditions, the proposed algorithm dynamically adapts to varying network conditions, providing a scalable and robust solution for FL in VEC.