Résumé : This paper proposes a resilient-backpropagation-neural-network-(Rprop-NN) based algorithm for Photovoltaic (PV) maximum power point tracking (MPPT). A supervision mechanism is proposed to calibrate the Rprop-NN-MPPT reference and limit short-circuit current caused by incorrect prediction. Conventional MPPT algorithms (e.g., perturb and observe (P&O), hill climbing, and incremental conductance (Inc-Cond) etc.) are trial-and-error-based, which may result in steady-state oscillations and loss of tracking direction under fast-changing ambient environment. In addition, partial shading is also a challenge due to the difficulty of finding the global maximum power point on a multi-peak characteristic curve. As an attempt to address the aforementioned issues, a novel Rprop-NN MPPT algorithm is developed and elaborated in this work. Multiple case studies are carried out to verify the effectiveness of the proposed algorithm.