Intelligent ANFIS-MPPT Control via Double-Diode Model Training: Implementation and Validation Under Tropical Irradiance Profiles in Cúcuta, Colombia

An adaptive neuro-fuzzy inference system (ANFIS)-based maximum power point tracking (MPPT) controller, trained with real-world data, is proposed and experimentally validated for photovoltaic systems operating under highly variable climatic conditions. Unlike conventional approaches relying on synthetic data, this work integrates a double-diode model (DDM) with high-fidelity measurements from a class-A pyranometer to capture the stochastic nature of the tropical environment. The controller was implemented on a Raspberry Pi 5 platform, governing a 100 kHz DC–DC converter connected to a 340 W photovoltaic array in Cúcuta, Colombia. Experimental results demonstrate a near instantaneous tracking time of 100 ms, significantly outperforming the 240 ms achieved by the conventional Perturb and Observe (P&O) method. Under rapid irradiance fluctuations, the system reached a peak static efficiency of 96.4% and a dynamic efficiency of 94.5%. Furthermore, a 10-h comparative energy analysis revealed that the ANFIS engine harvested 926.6 Wh, representing an 8.21% increase in cumulative energy yield compared to the P&O algorithm. This study confirms that intelligent MPPT controllers trained with real physical data can operate robustly on low-cost, non-deterministic embedded platforms, providing a resilient solution for energy optimization in modern PV microgrids.