I. Introduction
Global energy security positions solar energy as a strategic alternative, although the efficiency of photovoltaic systems is constrained by the non-linear nature of I-V curves under environmental variations. MPPT techniques are essential for optimizing this operation. However, conventional methods, such as P&O, fail under transient regimes. This work proposes the ANFIS model as a robust solution, integrating Artificial Neural Networks and Fuzzy Logic to provide dynamic adaptation without the need for complex mathematical modeling.

II. Methodology
An experimental database of 17,193 real-world irradiance samples, collected from a university plant, was utilized. The system was sized using Yingli polycrystalline modules (245 W, 15.1% efficiency) and simulated within the MATLAB/Simulink environment. The model was trained with 80% of the data and validated with the remaining 20%, exploring four scenarios that varied the number of membership functions (5 and 8) and training epochs (100 and 150). The ideal configuration was defined by the lowest prediction error and the highest coefficient of determination (?²).

III. Results
Simulations confirmed the high tracking capability of the ANFIS controller. It was observed that increasing the training to 150 epochs significantly optimized the stability of the estimated power. Despite initial oscillations typical of the transient regime, the voltage (?_??) and current (?_??) curves converged rapidly. Scenario B_2 (8 functions, 150 epochs) exhibited superior performance, achieving an ?² of 0.91 and the lowest Mean Squared Error (MSE) of 0.6895.

IV. Conclusions
The ANFIS model proved highly effective in extracting the maximum power point under real-world disturbances. The integration of climatic and electrical variables resulted in rapid responses and low oscillation around the MPP, validating the technique as a robust tool for the optimization of photovoltaic systems.