Energy management is one of the latest advances in the field of smart home technologies. Finding a way to save energy will help us to save money and reduce the emission of CO2 to the atmosphere, therefore preserving our environment. This study presents an improved architecture of long-short term meory (LSTM) to disaggregate individual appliance power signals from the overall power consumption of whole house. In order to reduce the training time, the proposed model has less parameters while the accuracy of estimation maintains the same or above that of the original LSTM model. The proposed deep learning structure has been tested on IAWE dataset. Experimental results show that relative error, mean absolute error and executing time have improved compared to the original LSTM structure.
Paper
Full text
Light-Weight Recurrent Deep Learning Algorithm for Non-Intrusive Load Monitoring
Semantic Scholar · Engineering · 2019
Abstract
Energy management is one of the latest advances in the field of smart home technologies. Finding a way to save energy will help us to save money and reduce the emission of CO2 to the atmosphere, therefore preserving our environment. This study presents an improved architecture of long-short term meory (LSTM) to disaggregate individual appliance power signals from the overall power consumption of whole house. In order to reduce the training time, the proposed model has less parameters while the accuracy of estimation maintains the same or above that of the original LSTM model. The proposed deep learning structure has been tested on IAWE dataset. Experimental results show that relative error, mean absolute error and executing time have improved compared to the original LSTM structure.