The digital transformation of electric power systems demands high-fidelity and real-time visualization of dynamic infrastructure scenes under diverse operational conditions. To address the challenges of high latency, poor adaptability, and dynamic complexity in traditional rendering approaches, we propose FMD-EdgeRender, a novel dynamic rendering framework based on cloud-edge-end collaboration. FMD-EdgeRender integrates foundation model-guided scene initialization, cross-scale temporal deformation modeling, and adaptive motion-aware interpolation to achieve robust, efficient, and real-time rendering of dynamic electric power scenarios. Extensive experiments on real-world collected datasets demonstrate that FMD-EdgeRender achieves superior rendering quality (38.7 PSNR), low latency (below 50 ms), and strong adaptability across a variety of operational environments. Our results validate the effectiveness of combining foundation models with dynamic deformation modeling for intelligent visualization in critical infrastructure systems.