Generative model-based hybrid forecasting model for renewable electricity supply using long short-term memory networks: A case study of South Korea's energy transition policy

Abstract

Forecasting renewable energy is essential for achieving a sustainable energy future. This study aimed to develop a hybrid deep-learning-based model for forecasting renewable electricity supply in a case study of South Korea in order to assist national-scale energy plan assessments. A generative model based on the variational auto-encoder (VAE) algorithm allows for the harnessing of numerous samples by solving problems such as a lack of sufficient time-series data and their uncertainties. Long short-term memory (LSTM) networks, which are well-suited to facilitate time-series problems, were used in forecasting, and they were compared to other machine learning-based models, such as the gated recurrent unit, deep neural network, and autoregressive integrated moving average models. Performance evaluation metrics, such as coefficient of determination (R2), root-mean-square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and weighted absolute percentage error (WAPE), were used to determine the optimal model. Therefore, the developed forecasting model based on the LSTM and VAE is suitable for the case study because it presented the highest R2 score of 0.92 and a decrease of 17%, 24%, 13%, and 14% in the RMSE, MAE, MAPE, and WAPE, respectively. It can be stated that results from this study will have a significant impact on renewable energy planning.