Deep Learning models for Smart Building Load Profile Prediction
Author(s)
Palak, M
Revati, G
Hossain, MA
Sheikh, A
Griffith University Author(s)
Year published
2021
Metadata
Show full item recordAbstract
With the development in the sector of smart cities, smart buildings are becoming quite popular. For effective energy management of smart buildings, the accurate prediction of the electrical consumption data is required. To increase the accuracy of prediction, the paper proposes deep learning methods for load profile prediction in a model-free environment. In this paper, a commercial smart building load profile is predicted using the re-current neural network (RNN), long short term memory (LSTM), bidirectional LSTM (BiLSTM), and gated recurrent unit (GRU) model. Using the best features of RNN and BiLSTM a hybrid model is ...
View more >With the development in the sector of smart cities, smart buildings are becoming quite popular. For effective energy management of smart buildings, the accurate prediction of the electrical consumption data is required. To increase the accuracy of prediction, the paper proposes deep learning methods for load profile prediction in a model-free environment. In this paper, a commercial smart building load profile is predicted using the re-current neural network (RNN), long short term memory (LSTM), bidirectional LSTM (BiLSTM), and gated recurrent unit (GRU) model. Using the best features of RNN and BiLSTM a hybrid model is developed for the load profile prediction. To compare the performance of different models error metrics such as Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) are considered for performance evaluation. The paper also highlights the significance of hyperparameters in improving the accuracy of prediction. Finally, from the results, it can be claimed that the hybrid model performs better as compared to RNN, LSTM, and GRU models.
View less >
View more >With the development in the sector of smart cities, smart buildings are becoming quite popular. For effective energy management of smart buildings, the accurate prediction of the electrical consumption data is required. To increase the accuracy of prediction, the paper proposes deep learning methods for load profile prediction in a model-free environment. In this paper, a commercial smart building load profile is predicted using the re-current neural network (RNN), long short term memory (LSTM), bidirectional LSTM (BiLSTM), and gated recurrent unit (GRU) model. Using the best features of RNN and BiLSTM a hybrid model is developed for the load profile prediction. To compare the performance of different models error metrics such as Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) are considered for performance evaluation. The paper also highlights the significance of hyperparameters in improving the accuracy of prediction. Finally, from the results, it can be claimed that the hybrid model performs better as compared to RNN, LSTM, and GRU models.
View less >
Conference Title
Proceedings of 2021 31st Australasian Universities Power Engineering Conference, AUPEC 2021
Subject
Electrical engineering