A Novel Digital Design Framework for Power Module Packaging Optimization Using Data-Driven Modeling and Intelligent Optimization Techniques

The miniaturization and high integration of power modules present significant challenges to conventional packaging technologies, particularly due to the increasing complexity of interconnections and parasitic effects. These challenges are especially pronounced during high-speed switching operations, where parasitic inductances can lead to significant voltage overshoots, oscillations, and electromagnetic interference (EMI). This study proposes a novel digital design framework for optimizing power module packaging, integrating data-driven modeling and intelligent optimization techniques. A random forest model is employed to accurately predict parasitic inductances, capturing complex nonlinear relationships, while a differential evolution algorithm is used for global optimization within a high-dimensional design space. The optimization results demonstrate a substantial reduction in parasitic inductance, from 14.57 nH to 9.04 nH, thereby enhancing the module’s performance during high-speed switching, reducing voltage overshoot and oscillation, and improving electromagnetic compatibility and overall reliability.