Unknown Load Detection for Nonintrusive Load Monitoring Based on Feature Embedding and Nearest Neighbor Learning

Non-intrusive load monitoring (NILM) can effec-tively evaluate and track fine-grained energy consumption at the appliance level by using instruments or devices such as smart meters, providing a reasonable and safe energy measurement solution. However, existing NILM algorithms rely on closed-set recognition, which cannot detect the continuous addition of new home appliances in the real world. Moreover, conventional NILM methods face challenges in feature selection, such as in-formation loss and ambiguous distinguishing features, which further reduces the recognition rate on the measuring device. To address unknow load detection, this paper proposes a feature embedding and nearest neighbor learning method using Dual-channel-LUNAR. The dual-channel network first leverage fea-ture embeddings extracted from original signal to reduce the intra-class distance for known devices, thereby distinguishing them from untrained unknown devices. LUNAR then detects unknown devices by employing nearest neighbor distance learn-ing in the embedded feature space. Furthermore, to improve the distinction between appliances with similar running modes in feature selection, a new kind of three-channel load signature is generated by constructing Relative Position Matrix, V-I trajecto-ry, and Mel-spectrogram. Lastly, for feature learning, the dual-channel network consists of a residual convolutional network embedded with Coordinate Attention for learning multi-level semantic features and a transformer network along with RoPE for better adaptation to current sequence tasks. Comprehensive experiments conducted on two public datasets highlight the ad-vantages of our proposed approach.