Artificial neural network–based phase-shift optimization for input ripple minimization in an interleaved boost–Z-source converter

This research introduces a neural-network-based control strategy to reduce input current ripple in an asymmetric interleaved Boost–Z-source converter. Unlike traditional fixed-phase interleaving that only reduces ripple at certain operating points, this method adjusts the interphase shift based on the converter's immediate duty cycles.

Researchers first created a ripple-surface map over a wide range of operating conditions, determining the optimal phase shift for each duty-cycle combination to minimize RMS ripple. They then trained a compact 2-10-1 feedforward artificial neural network (ANN) with these optimal points to predict the phase shift needed for ripple minimization in real time.

This ANN was implemented on a cost-effective STM32 microcontroller and integrated into the converter's PWM generation system. Concrete experiments on a full-scale prototype confirmed the simulation results, proving the method's effectiveness and reliability. The proposed technique merges the precision of data-driven modeling with the ease of embedded implementation, presenting a promising substitute for traditional ripple-minimization techniques in asymmetric or hybrid power converter designs.

It exhibits rapid response to duty-cycle changes and aligns well with both simulation and hardware outcomes. This work was conducted by researchers from the Department of Electrical Engineering at Azarbaijan Shahid Madani University in Tabriz, Iran. The study is licensed under a Creative Commons Attribution 4.0 International License, allowing for the reuse of the material with proper attribution.