Machine Learning Methods for Power Flow Control of Multi-Active-Bridge Converters

Controlling the multiway power flow in a multi-active-bridge (MAB) converter is important for maintaining high performance and realizing sophisticated functions. Traditional methods for MAB converter control rely on precise circuit models of the system. This paper presents machine learning (ML) methods for controlling the power flow in a MAB converter. A feedforward neural network (FNN) was developed to capture the non-linear characteristics and predict the phases needed to achieve the targeted power flow. The neural network was trained using a large amount of data collected with a set of known phase angles, and was used to predict the phase to achieve the targeted multiway power flow. A 6-port MAB converter was built and tested to verify the theory, validate the methodology, and demonstrate the hardware implementation of the ML-based control strategy. Machine learning based power flow control methods can achieve comparable accuracy as traditional model-based power flow control methods, and can function without a precise lumped circuit element model of the MAB converter.