Dublin Core
Title
Solar Irradiation Prediction Based on M5 Model Tree and Feature Importance Evaluation
Abstract
In the last decade, the usage of renewable energy is on the rise, and that trend will only continue because technology is becoming more developed, so renewable energy sources are going to offer more for the same price. Besides all positive properties, there
are also some negatives like direct dependence on the weather conditions. That means energy production is constantly changing, so it must be as precisely as possible predicted to be usable on a large scale. Fifteen attributes were analyzed using M5 regression tree.
High positive degree of correlation was found between participle water and dew point temperature, air temperature with dew point, air temperature with precipitation of water, snow depth with Albedo daily, zenith angle with relative humidity, GHI with Air temperature. It was found that the zenith angle, between the normal of the Earth’s surface and the Sun, was the most important feature of the dataset for solar irradiation prediction.
are also some negatives like direct dependence on the weather conditions. That means energy production is constantly changing, so it must be as precisely as possible predicted to be usable on a large scale. Fifteen attributes were analyzed using M5 regression tree.
High positive degree of correlation was found between participle water and dew point temperature, air temperature with dew point, air temperature with precipitation of water, snow depth with Albedo daily, zenith angle with relative humidity, GHI with Air temperature. It was found that the zenith angle, between the normal of the Earth’s surface and the Sun, was the most important feature of the dataset for solar irradiation prediction.
Keywords
machine learning, photovoltaic, renewable energy, solar irradiation, weather forecast.
Identifier
ISSN 2637-2835 (Print)
DOI
10.14706/JONSAE2022422
Publisher
International Burch University
Language
English language
Type
Original research