Long Short-Term Memory Decomposition Model for Forecasting Solar Irradiance Photovoltaic Household Scale on Grid

A proposed study using the development of the Long Short Term Memory Decomposition hybrid method for maximalization forecasting of solar irradiance (SI) based on meteorology data using a development decomposition and long short-term memory model. This combination modeling is a development decomposition model and an LSTM model. The combination development decomposition long short-term memory (DD-LSTM) modified hybrid model is designed to predict solar irradiance photovoltaic household scale on grid 1 hour ahead based on weather data for solar irradiance household scale on grid photovoltaic station. The proposed research of this hybrid model is taken into calculation based on meteorology data. The method development decomposition is a process of training and testing to the Long Short Term Memory model. In the error statistical calculation index of the DD-LSTM model, the mean absolute percentage error (MAPE) is 1 %. The hybrid combination model forecasting is then compared to actual data, development decomposition, and training results indicate that the DD-LSTM model explained in the study can calculate hourly solar irradiance with optimal accuracy.