Speed forecasting has become a critical component in smart transportation systems, particularly for optimizing energy management in electric vehicles. While existing forecasting methods show promise, they often face significant limitations when deployed in IoT-enabled smart systems environment. In this paper, we introduce IoT-Enabled FM2I, an IoT-optimized approach for EV speed forecasting by enhancing the Forecasting Method by Image Inpainting (FM2I) method through a hinge selection strategy and edge computing. Our method enhances the traditional FM2I by adapting the hinge selection process for forecasting. The approach transforms time-series data into image representations, applies a mask depending on the given forecast horizon, and employs an enhanced patch-based image inpainting technique. We evaluate this approach using real-world data collected across various urban environments. Results demonstrate that HS-FM2I achieves better accuracy compared to traditional methods while maintaining real-time performance on embedded devices.