<![CDATA[General Background: The increasing demand for electrical energy has accelerated the development of renewable energy systems, particularly photovoltaic (PV) technology, which requires reliable power estimation under varying environmental conditions. Specific Background: PV output power is strongly affected by environmental parameters such as light intensity, voltage, current, and temperature, making prediction difficult when nonlinear relationships occur among variables. Knowledge Gap: Conventional multi-regression approaches have limitations in modeling nonlinear PV characteristics, while comparative evaluations of Adaptive Neuro-Fuzzy Inference System (ANFIS) configurations for PV power estimation remain limited. Aims: This study aims to develop and evaluate an ANFIS Sugeno-based model for estimating PV output power and compare its performance with multi-regression methods using real-time environmental data collected through an Arduino-based data logger. Results: The developed data logger successfully recorded stable real-time data, while the ANFIS model demonstrated substantially lower prediction errors than multi-regression. The best-performing configuration, Gauss555, achieved Mean Absolute Percentage Error (MAPE) values of 2.03% for training data and 2.13% for testing data, whereas multi-regression produced errors of 54.63% and 79.19%, respectively. Gaussian membership functions consistently generated lower and more stable Absolute Percentage Error (APE) values than triangular and trapezoidal functions. Novelty: The study integrates real-time PV environmental monitoring with comparative ANFIS membership function configurations to identify the most suitable nonlinear prediction model for PV output estimation. Implications: The findings demonstrate that ANFIS provides a robust and accurate approach for photovoltaic power estimation, supporting reliable renewable energy management and future intelligent PV monitoring systems.]]>
