Abstract
<title>Abstract</title> <p>Electric-vehicle (EV) powertrains supplied from low-voltage battery packs demand power-conditioning stages capable of establishing a high and tightly regulated DC-link while preserving efficiency, dynamic agility and long-term reliability. Conventional boost and quadratic-boost converters either provide insufficient gain or operate at extreme duty ratios that degrade efficiency and stress semiconductors. This paper presents a split-capacitor-type elementary additional-series positive-output super-lift DC–DC converter that elevates an 8 V source to a regulated 42 V DC-link, together with an artificial-intelligence-assisted model-predictive-control (AI-MPC) scheme and a comprehensive reliability assessment. The converter employs a single inductor, two controlled switches, nine diodes and five split capacitors to realise a geometric voltage-lift mechanism, achieving a conversion ratio of 5.25 at a nominal duty of 0.5 while confining the device voltage stress. A complete continuous- and discontinuous-conduction-mode analysis is derived from the inductor volt-second and capacitor charge-balance principles, and is extended to a state-space averaged model and a small-signal formulation that exposes the right-half-plane zero limiting the achievable bandwidth. The proposed controller augments a finite-horizon predictive core with a feed-forward artificial neural network that maps the battery voltage, load power, inductor current and output voltage to an optimal duty command, reducing the online optimisation burden and improving transient recovery. A reliability framework based on the MIL-HDBK-217F part-stress methodology yields component failure rates, the system mean-time-between-failures and the time-dependent reliability function, while a Monte-Carlo study with 1000 parameter realisations quantifies robustness to component tolerances. Simulation and hardware-oriented results demonstrate that the AI-MPC scheme reduces settling time and overshoot by more than 60% and 78% respectively relative to a proportional-integral baseline, sustains a peak efficiency of 96.5%, and improves the predicted reliability margin, confirming the suitability of the converter for demanding EV applications.</p>