To stably operate EMS(Energy Management System) that systematically manages energy in the building, there is a need for predicting energy consumption and production in advance. In this study, we used hourly observed meteorological data from January, 2...
To stably operate EMS(Energy Management System) that systematically manages energy in the building, there is a need for predicting energy consumption and production in advance. In this study, we used hourly observed meteorological data from January, 2018 to December, 2020 provided by the Seoul branch of the Korea Meteorological Administration. The solar altitude data are calculated through the solar angle calculator provided by One Energy. The collected data are converted into available data set by data processing strategy. The correlation analysis between each meteorological data type and solar radiation after one hour proceeded to select the input parameters on the developed model. Selected meteorological data sets are used in the learning stage of the developed LSTM structure prediction model. The predictive performance of each model were analyzed through MAE(Mean Average Error), NMBE(Normalized Mean Bias Error), CV(RMSE)(Coefficient of Variation of Root Mean Square Error), R²(Coefficient of Determination) and computational time. The model with a window size of 24 was selected by performance evaluation criteria. Valid predictive performance of solar radiation after one hour in Busan was derived also from the selected model.