Physical Mechanism and Data Hybrid Modeling for Device-Level Emulation of Modular Multilevel Converter

In the modular multilevel converter (MMC), the insulated gate bipolar transistor (IGBT) is one of the most important devices for power conversion. Therefore, an accurate and efficient model of the sub-module is significant for design, monitoring and diagnosis of valve towers. Due to the complex multi-physics field inside the semiconductor device, it is laborious to describe the physical mechanism model in detail. Traditional mechanism models cannot meet the requirements of digital twin applications. In this paper, a new modeling method for sub-modules based on long short-term memory neural network is developed and the model is integrated into the traditional model of MMC for simulation, which is a physical mechanism and data hybrid modeling method to retain both of their advantages.