An Approach Using Transformer-based Model for Short-term PV generation forecasting
Quoc‐Thang Phan, Yuan‐Kang Wu, Quốc Dũng Phan
Abstract
Solar power forecasting has already become a key role in energy market. However, forecasting PV generation is a challenging task because solar energy strongly depends on weather conditions, day/night cycles and meteorological variables such as solar irradiance, temperature, humidity, etc. Therefore, integrating solar power into grids requires accurate prediction methods, which is also improve the quality of the power systems and reduce operational cost. Beside traditional approaches for forecasting solar PV power generation, new techniques arise every year to enhance the performance of models with the main objective to reduce uncertainties. In this paper, a novel deep learning model based on transformer has been established for one-hour-ahead PV generation forecasting. In practice, this work uses PV power output data with one-hour resolution from the North PV sites in Taiwan, and 2 years of Numerical Weather Prediction (NWP) data from Central Weather Bureau (CWB). Furthermore, forecasting results show that the proposed model outperforms other models such as Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) in terms of NRMSE and NMAPE.