| Abstrakt: |
Power generation from renewable energy sources is key to a clean energy future and solar energy is world's fastest growing energy sector. Solar energy is renewable, CO2-free and with low operational cost. There are several advantages of using solar energy; however, it does have a few drawbacks such as hefty initial cost, high-priced storage, weather dependency, sizable space requirement, etc. As such, it is critical to predict solar radiation (SR) in an accurate and efficient way to install solar plants in optimal locations. Factors like global horizontal irradiance (GHI), temperature, humidity, cloud cover, wind speed, etc. make SR highly intermittent and variable. Accurate forecasting of SR is vital to finalize installed capacity of the proposed power plant but is extremely challenging due to unpredictability of sunlight. Even the world's best organizations such as the International Energy Agency (IEA) finds it difficult to accurately predict SR. In this work, we analyze global and diffuse SR data gathered from India Meteorological Department (IMD) Pune. This data is first analyzed for features, dominant features are identified, and several machine learning algorithms are employed. Random forest method provides the best results on this data based on several quantitative measures. [ABSTRACT FROM AUTHOR] |
