Load Profile Prediction in Smart Building using Data Driven Approaches

With the emergence of the smart grid and smart city trends, smart buildings are gaining attention. Energy management is a key component of a smart building management system, and it necessitates precise forecasting of the building's electrical energy usage profile. In view of this, the paper focuses on the analysis of a data-driven approach for predicting the electricity consumption of a smart building in a model-free environment. For predicting the load profile the dynamic mode decomposition (DMD) and deep learning models such as recurrent neural network (RNN), long short term memory (LSTM), bidirectional LSTM (BiLSTM), and gated recurrent unit (GRU) model are considered in this paper. The paper also proposes the development of a hybrid model using the best features of RNN and BiLSTM. A comparative study is carried out for analyzing the performance of different data-driven approaches and for that various test scenarios were conducted to evaluate the prediction accuracy. Finally, from the prediction results, it can be claimed that the deep learning models especially the proposed hybrid model outperform other deep learning methods as well as DMD due to the consideration of hyperparameters.