Pemodelan Energi Listrik yang Dihasilkan oleh PV Menggunakan Metode Time Series dan Neural Network untuk Komparasi

Abstract

Renewable energy sourced from the sun has become one of the focal points of alternative renewable energy as fossil energy reserves diminish. Solar energy, which is converted into electricity using photovoltaic technology, is influenced by several variables, particularly weather variables such as temperature, humidity, and solar radiation. This study involves modeling and forecasting the power output of a 100 Watt PV Solar system using Time Series Analysis and Neural Network techniques. The PV solar system is connected to various weather variable measurement sensors, such as a pyranometer, temperature sensor, and humidity sensor. The data collected from these sensors serve as input for calculating the power output of the installed 100 Watt PV system. The power output is observed on an hourly and daily basis. The modeling results indicate that the best model obtained using ARIMA with variables is ARIMA (0,0,2), incorporating all weather variables (Radiation, Humidity, Temperature*, Wind, and Light*) with a MAPE (Mean Absolute Percentage Error) of 2.91%. Meanwhile, for the best Neural Network (LSTM) model, the input variables of radiation, temperature, and intensity achieved a MAPE of 3.41%