Open AccessMeteorological Factor Multivariate Models Affecting Solar Power Prediction using Long Short-term Memory
Author(s) -
Nam Rye Son,
Si Yang
Publication year - 2020
Publication title -
international journal of innovative technology and exploring engineering
Language(s) - English
Resource type - Journals
ISSN - 2278-3075
DOI - 10.35940/ijitee.d1172.029420
Subject(s) - cloud cover , meteorology , environmental science , humidity , wind speed , solar energy , sunshine duration , term (time) , solar power , computer science , cloud computing , power (physics) , engineering , relative humidity , electrical engineering , geography , physics , quantum mechanics , operating system
Solar power systems have been recently installed in buildings to efficiently manage their energy consumption and production in them. Because electrical energy is produced and consumed simultaneously owing to its physical nature, it is necessary to predict the exact solar power necessary to maintain a stable power supply. To manage the building energy management system (BEMS) effectively, this paper proposes 6 models (solar radiation, sunlight, humidity, temperature, cloud cover, wind speed) and compares the performances of these models. Through this comparison, we solved the traditional long short-term memory (LSTM) problem and proposed a new LSTM. It was determined that the meteorological factors for forecasting solar power varied by season. The performance was shown in order of solar radiation, sunshine, wind speed, temperature, cloudiness and humidity at annual average. Additionally, the proposed LSTM performed better than the traditional LSTM.