Multi–horizon energy consumption forecasting on daily basis

Forecasting of energy consumption has crucial role in stability of energy systems. Furthermore, the need of forecasting energy consumption has drastically raised with the energy crisis we faced and with fast development of technology, especially artificial intelligence. Public institutions work hard to drive faster to achievement of sustainable energy. All this exposes the need to develop deep-learning model with attention for forecasting energy consumption on daily basis. CNN-LSTM model with attention layer is proposed to forecast energy consumption on three different time horizons 7, 14 and 21 days ahead. Model is tested on real energy consumption data of public educational institution Faculty of Electrical Engineering and Information Technologies in Skopje. The achieved results show great performance of model in multi-step forecasting.