Final published version
Research output: Contribution to Journal/Magazine › Journal article › peer-review
Research output: Contribution to Journal/Magazine › Journal article › peer-review
}
TY - JOUR
T1 - Improving wireless indoor non-intrusive load disaggregation using attention-based deep learning networks
AU - Liu, Qi
AU - Zhang, Jing
AU - Liu, Xiaodong
AU - Zhang, Yonghong
AU - Xu, Xiaolong
AU - Khosravi, Mohammad
AU - Bilal, Muhammad
PY - 2022/4/30
Y1 - 2022/4/30
N2 - The intensification of the greenhouse effect is driving the implementation of energy saving and emission reduction policies, which lead to a wide variety of energy saving solutions benefiting from the advancement of emerging technologies such as Wireless Communication, the Internet of Things, etc. With the multi-convergence development of different domains in the power industry, demand-side refinement management solutions are constantly concerned. One of the key functions of demand-side refinement management solutions is non-intrusive load monitoring (NILM), which has benefited from the growing interest in emerging technologies such as wireless communications and the Internet of Things. Currently, deep learning methods such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) are widely used for in-depth research on NILM. This paper investigates the role of attention mechanisms in the above two time-series deep learning models. Experiments show that the improved model is more than 10% more effective in indoor scenes, especially for typical household appliances such as refrigerators.
AB - The intensification of the greenhouse effect is driving the implementation of energy saving and emission reduction policies, which lead to a wide variety of energy saving solutions benefiting from the advancement of emerging technologies such as Wireless Communication, the Internet of Things, etc. With the multi-convergence development of different domains in the power industry, demand-side refinement management solutions are constantly concerned. One of the key functions of demand-side refinement management solutions is non-intrusive load monitoring (NILM), which has benefited from the growing interest in emerging technologies such as wireless communications and the Internet of Things. Currently, deep learning methods such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) are widely used for in-depth research on NILM. This paper investigates the role of attention mechanisms in the above two time-series deep learning models. Experiments show that the improved model is more than 10% more effective in indoor scenes, especially for typical household appliances such as refrigerators.
KW - Attention
KW - CNN
KW - Load management
KW - LSTM
KW - NILM
U2 - 10.1016/j.phycom.2021.101584
DO - 10.1016/j.phycom.2021.101584
M3 - Journal article
AN - SCOPUS:85122804545
VL - 51
JO - Physical Communication
JF - Physical Communication
SN - 1874-4907
M1 - 101584
ER -