Optimized Ripple Prelearning Process for Phase-Shift Control of Interleaved Multiphase DC–DC Converters

Minimizing ripples and harmonics is an important criterion in multiphase dc–dc converters. In case of physical parameters degradation, nonidentical phases, or fault occurrence, the symmetrical interleaved mode ceases to be the optimum mode, and a supervising system is required. This article proposes a novel method for optimizing the current ripples and harmonics in multiphase dc–dc converters based on asymmetrical phase shifting using only the sensed variables by the control loops. The proposed method uses a developed online phases current shape-factors (PCSFs) calculation algorithm based on the currents derivative with an artificial neural network (ANN), offline trained, using an enhanced particle swarm optimization (PSO) technique on a high-fidelity model of the converter. The goal is to meet the desired current requirements by adapting phases shifting dynamically during the converter's operation at each period. Moreover, in the case of an open circuit fault (OCF), the proposed method can successfully adapt the phase shifting between the remaining healthy phases, reducing the current ripples caused by the fault. The proposed technique is validated experimentally using a three-phase interleaved boost converter, illustrating the current improvement in various scenarios such as OCFs and nonidentical parameters, proving an effective phase-shifting adaptation against uncertainties and variations in the parameters.