Comparison of solar panel energy production forecast with machine learning methods

Countries have turned to cheap and reliable renewable energy production due to the increase in electricity energy needs. Photovoltaic (PV) energy systems have been a prevalent one among these energy resources in recent years. In the estimation of generated solar energy using machine learning methods, meteorological data like solar radiation, pressure, and temperature are influential elements. The future generation of energy in solar power plants can be quite difficult to predict since these elements vary depending on weather conditions. Methods have been developed to calculate potential electricity generation from radiation energy before the installation of plants. One of these methods is to create a model and obtain simulation results by using machine learning algorithms that have become widespread recently. In this thesis study; It was aimed to predict the hourly electric power generation in the photovoltaic panels in Amasya University, Ipekkoy campus, using machine learning models. The effects of meteorological factors on energy production were analyzed. Machine learning was carried out using the libraries of the python programming language for the evaluations. In the study, temperature, solar radiation, azimuth angle and produced energy data of 2016 Amasya region were used as data set. 80% of the data in the study were used for the learning dataset, while 20% were used for testing data. For data learning, machine learning algorithms of Linear Regression models, Support Vector Regression, AdaBoost Regressor, K-Nearest Neighbors, Random Forest, Naive Bayes, and Artificial Neural Network were used. As a result, the solar energy potential was successfully estimated using the obtained data.