ANFIS Design Based on Prediction Models for The Photovoltaic System

Abstract

Photovoltaic system has uncertain output in generating electrical energy, as it is intensely influenced by different weather condition. This study discusses the photovoltaics model to predict the output power precisely. This modeling system applies an Adaptive Neuro-Fuzzy Inference System (ANFIS) technique to gain data of power prediction, voltage, current, and temperature. The mathematical representation of the photovoltaic using Matlab/Simulink setting has been developed and presented by using the photovoltaic basic sequence equation, including solar irradiation effect and temperature changes. This model is divided into two systems run by ANFIS; ANFIS 1 and ANFIS 2. The design of ANFIS is expected to update its parameter to determine errors between output and target. MAPE (Mean Absolute Percentage Error) value for ANFIS 1 test of open circuit output voltage was 0.0104. This MAPE score is found to be excellent predictive data with less than 10% MAPE value. For the ANFIS 2 test, the AC output voltage was 0.026%, output current of 1.3035%, and 0.0046% of frequency. Based on the MAPE scores, very suitable data prediction has been produced with less than 10% MAPE value. Briefly, this study reveals that the ANFIS technique yields load prediction results that can improve the accuracy and rapidness of prediction as well as very minimum errors.