Forecasting the Global and Diffuse Solar Irradiance of Bangkok's Tropical Climate using Long-Short Term Memory (LSTM) Technique

This paper presents the model development to forecast the one-hour ahead global and diffuse solar irradiance of Bangkok, Thailand, located in a tropical climate zone. Based on the measurement records during 2019–2021, the forecasting models were developed by the long-short-term memory (LSTM) technique in which the ambient temperature, solar elevation, and an insolation parameter namely normalized global solar radiation were selected as the model inputs. To assess the forecasting performance, the LSTM models were compared to the artificial neural network (ANN) model and the simple persistence (SP) model. Two statistical parameters of normalized mean bias error (nMBE) and normalized root mean square error (nRMSE) were used for the performance assessment. The results showed that the LSTM model was superior to the other two models (ANN and SP) for all weather conditions (both wet and dry periods). The LSTM model also forecasted the global irradiance more accurately than the diffuse irradiance.