Physics-Infused Neural MPC of a DC-DC Boost Converter with Adaptive Transient Recovery and Enhanced Dynamic Stability
This work addresses control challenges in power electronics, offering a computationally efficient and physically consistent solution for real-time applications, though it appears incremental as it combines existing methods like PINN and FCS-MPC.
The paper tackled the problem of controlling DC-DC boost converters for efficiency and stability under varying loads by proposing a hybrid physics-informed neural network combined with finite control set MPC, resulting in improved transient response, reduced voltage ripple, and robust operation in experiments on a commercial module.
DC-DC boost converters require advanced control to ensure efficiency and stability under varying loads. Traditional model predictive control (MPC) and data-driven neural network methods face challenges such as high complexity and limited physical constraint enforcement. This paper proposes a hybrid physics-informed neural network (PINN) combined with finite control set MPC (FCS-MPC) for boost converters. The PINN embeds physical laws into neural training, providing accurate state predictions, while FCS-MPC ensures constraint satisfaction and multi-objective optimization. The method features adaptive transient recovery, explicit duty-ratio control, and enhanced dynamic stability. Experimental results on a commercial boost module demonstrate improved transient response, reduced voltage ripple, and robust operation across conduction modes. The proposed framework offers a computationally efficient, physically consistent solution for real-time control in power electronics.