Residual Seq2Seq model for Building energy management

According to the UN Department of Economic and Social Affairs (DESA), the global urbanization rate is expected to reach 68% in 30 years from 55% in 2018. As global urbanization progresses, the proportion of total energy consumption consumed in buildings is increasing, and efficient energy management in buildings is becoming increasingly important in terms of efficient use of global natural resources and air quality management. This paper presents an implementation detail of the energy consumption prediction deep learning model for efficient building energy management. The developed model is multilayer LSTM seq2seq model which predicts energy consumption for one day by using environmental data and energy consumption data measured in a real testbed. In a multi-zone building, one zone of data is used, which is characterized by the fact that data patterns over time are not neatly repeated. This paper presents deep learning depth and performance changes as adding layers of seq2seq model in building energy consumption forecasting. In addition, it presents performance comparison with other algorithms.