The advancement of green power grids demands enhanced performance from core power electronics like Insulated Gate Bipolar Transistors (IGBTs), crucial for grid stability. While existing digital twin (DT) models aid IGBT optimization, limitations in accuracy, adaptability, and data processing hinder high-precision prediction for practical grid applications. This study proposes a novel IGBT chip performance optimization method using next-generation DT technology. This DT leverages multi-physics coupling modeling to capture IGBT's complex environments and operating modes, alongside real-time data interaction for model updating. A key innovation is the deep integration of data assimilation algorithms, uniquely tailored for IGBTs, to correct model errors by combining observational data with DT predictions. Performance optimization experiments under various conditions, combining multi-physics simulation and data assimilation, demonstrate that this method significantly improves prediction accuracy compared to traditional approaches. The optimized IGBTs exhibit superior power conversion efficiency, thermal management, and disturbance resistance, enhancing grid stability and security. This work highlights the significant potential of next-generation DT technology, augmented by data assimilation, in advancing intelligent green grid development, improving energy efficiency, and reducing operational costs.