A comparative study of different deep learning models for mid-term solar power prediction
R. Sivanand, Jay Singh, Abdul Ali, Aarman Khan, Md Fahimul Hoda
Abstract
Solar energy is an abundant and renewable form of energy that is the answer to the increasing energy requirements of today's world. Since the prediction of solar radiation shows some uncertainties depending on atmospheric parameters, a significant method of solar power forecasting for small and big businesses is necessary. In this paper based on Deep Learning algorithms, we put forward the models which can be employed to predict solar PV energy output in the energy industry. In this study, a comparison of different deep learning techniques is proposed. On average, we find that the models LSTM (Long short-term memory) and GRU (Gated recurrent unit) perform well. RNN, LSTM and GRU have been widely used in solar power forecasting. And their hybrid models have also improved their performance. The results show a comparative study of the three aforementioned models.