Application of Artificial Intelligence Techniques to Enhance the Solar Power Generation: Case Study in Sri Lanka

The increasing energy demand, high fuel import cost, air pollution and global warming are indicating the need of elimination of fossil fuels and it also have heightened the quest for cleaner and renewable energy sources. Even though the unique climatic and geographic characteristics of Sri Lanka pose opportunities for solar power generation, the country is struggling to optimize the full potential of solar power resource due to the less technological involvement. Since solar energy variable with the unpredictability of nature, determining the future solar power generation in advance which is called solar forecasting will give the solar energy sector with a great advantage. The previous studies used machine learning algorithms to anticipate the solar power that have shown greater results than traditional forecasting methods. This research involves the impact of solar module temperature in solar panel efficiency and the implementation of Long Short-Term Memory (LSTM) model for solar module temperature prediction. By using the predicted module temperature and other environmental parameters such as solar irradiance, windspeed and ambient temperature, the LSTM model can predict the accurate solar power forecasting for one day time horizon. Experimental results show the difference in accuracy level when using the predicted module temperature for forecasting than using actual module temperature.