KEYNOTE SPEAKERS

Andrew, Chen-Yeon CHU Ph.D. 

Professor, Feng Chia University 

Director, Institute of Green Products, Feng Chia University 

Executive Secretary, APEC Research Center for Advanced Biohydrogen Technology

Title: Deep Learning Model Established Using Satellite Cloud Images for Short-term Power Generation Forecasting of Solar Power Plants

As the demand for solar power generation increases, the forecast of solar energy becomes more and more important. Solar photovoltaic system is the main source of solar power. Solar power plants’ management and operation with energy storage systems need reliable solar power generation forecasts. These power plants can work like conventional power plants, avoid the uncertainty of power output and can compete in the energy market. Solar power generation prediction technology will have a significant impact on the future of large-scale solar energy plants. Forecast the power generation of solar photovoltaic systems dramatically depends on Climatic conditions, which will fluctuate over time.

This research is based on the historical data of the solar power generation systems of the Ming-Lun Senior High School, Taipei City and Shu Guang Girl’s Senior High School, Hsinchu City in Taiwan, and Asian Development College for Community Economy and Technology, Chiang Mai Rajabhat University in Thailand. The data also collected from Japanese himawari-8 satellite cloud images and the parameters of the position of the sun relative to the power plant location. The cloud image uses CNN to extract features and LSTM to forecast the next set of features. The feature selection is performed by comparing the direction vector X and Y of the sun relative to the power plant location, and the prediction results are corrected according to the angle between the direction of sunlight and the solar panel. A deep learning model for prediction solar power generation based on wide-range cloud obscuration rate was established. Use deep learning methods to train the model to forecast power generation. The final forecast results can be obtained that the revised forecast accuracy rates of the Chiang Mai case in Thailand, Hsinchu city case and Taipei city case were 92%, 90% and 85%, respectively. Finally, it was found that the Thailand case’s revised forecast accuracy rate was higher than 90% of the general revised forecast accuracy rate of the single international case.