Prediction of photovoltaic energy from augmentation of local temperatures based on machine learning algorithm

Abstract

As primary source for energy generation, consumption of fossil fuel has steadily increased along with growing demand for the energy. As a result, potential risks concerning environmental issues have surfaced due to high amount of carbon dioxide occurred during process of burning fossil fuel. To mitigate the growing concerns for the environment, renewable energy resources are introduced for their eco-friendly process and sustainability. Especially, consumption of solar energy has noticeably increased over past 10 years. However, its penetration into power system and market has brought new issues to those involved. Mainly, their issues originate from large uncertainty of photovoltaic energy generation. Since photovoltaic energy are generated based on sunlight exposed on solar panel, its generation largely depend on weather condition around the panel. In recent studies, they take advantage of weather forecast to predict future generation of photovoltaic energy otherwise utilize past data of the generation for the prediction. However, their approaches are limited in that their consideration for local weather condition are excluded. Therefore, in this paper, local weather information are incorporated with meteorological information from weather station for the prediction. As local weather information are not available from weather forecast, this paper initially proposes a method to extract the local information from weather forecast. Then, prediction of photovoltaic energy generation are proposed using combination of the extracted local data with weather forecast. The simulation result using real world data shows that the proposed method of photovoltaic energy prediction achieves 18% reduction of error rate compared to conventional approach that only depend on meteorological information in weather forecast.