Deep Learning and Federated Learning for Non-Intrusive Appliance Load Monitoring (NIALM)

Title: Deep Learning and Federated Learning for Non-Intrusive Appliance Load Monitoring (NIALM) Author: Huan Chen Affiliation: National Chung Hsing University, Taiwan Abstract: Non-Intrusive Appliance Loading Monitoring (NIALM) is a type of power internet of things (PIoT) which can perform energy dis-aggregation for home appliances in use. NIALM is a cost-effective solution due to less power meter should be installed, but it is a very challenging research due to its complexity when multiple appliances are in use at the same time. In recent years, with the explosive growth of artificial intelligence (AI) technologies, many advanced algorithms based on deep learning are proposed with very promising results. As such, this talk will first briefly review some important development of the NIALM research, then we will introduce our proposed algorithm called CAEBN as an example to show how to develop NIALM using deep learning. The NIALM problem is first modeled as a regression problem and we will show that the proposed method can predict the target signal correctly. The proposed CAEBN is designed based on the one-dimensional Convolutional Neural Networks (1D-CNN) Autoencoder with batch normalization. 1D-CNN autoencoder is used to extract the temporal features, while the BN is used to re-adjust the output distribution of each layer to prevent the gradient vanishing or explosion problem in the training process. In order to take the advantages of modern deep learning algorithms, in this talk, we will share our experience on the neural network training and hyper-parameter optimization issues. The last part of the talk will introduce recent emerging technology called federated learning for NIALM, which considers the usage scenario with privacy concerns for the electricity consumers. In this talk, we will extend our work to the novel federated learning framework. Some development tools and resource we use to implement the federated NIALM will be introduced. We will also compare the results with the conventional ones. At last, the trade-off between accuracy and privacy, and some other open research issues will be discussed.

Видео Deep Learning and Federated Learning for Non-Intrusive Appliance Load Monitoring (NIALM) автора Изумительные ресницы для изумительного образа