Multilabel Appliance Classification With Weakly Labeled Data for Non-Intrusive Load Monitoring

Using explainability tools to inform NILM algorithm performance

“…Table 1 compares F-SCORE performance of our DT multiclassifier with the following NILM multi-classifiers for Kettle (KE), Toaster (TOA), Washing Machine (WM), Microwave (MW) and Dishwasher (DW): (a) DT of [4] using 𝐸𝐷𝐺𝐸_𝑃 and 𝐸𝐷𝐺𝐸_𝑁 features and tested with REFIT House 2 (October 2015), (b) DT of [7] using 𝐸𝐷𝐺𝐸_𝑃, 𝐸𝐷𝐺𝐸_𝑁 and active power as features, tested on REFIT House 2 (October 2014), (c) deep learning multi-classifiers, LSTM, Convolutional Recurrent Neural Networks (CRNN, S-CRNN) and Semi-Supervised Multi-Label TCN (SSML-TCN) whose results with 100% strong labels are reported in [15], trained on appliance activations and tested on unseen REFIT Houses 4, 9 and 15. Best accuracy score is indicated in bold for each appliance, showing that the proposed approach has comparable performance w.r.t other state-of-the-art multiclassifiers in the literature.…”

Knowledge Distillation for Scalable Nonintrusive Load Monitoring

“…Recently, in [11], a semisupervised KD approach has been proposed to improve the transferability on target environments. Differently, in [7], the authors adopt a weakly supervised multilabel approach to reduce the labeling effort to train a CRNN, using both weakly labeled data (labels provided for a group of consecutive samples, e.g., for a 4 hours period) and strongly labeled data (i.e., labeled sample-by-sample). Successively, a transfer learning approach based on weak labels has been proposed in [13].…”

“…Due to the availability of a large quantity of low-frequency electrical load measurements from smart meters, deep learning (DL) approaches have recently become popular, representing the current state of the art in NILM both for regression and classification tasks [5], [6], [7], [8], [9], [10], [11], [12], [13].…”

“…Transfer learning typically requires collecting new data directly from the target environment to fine-tune pretrained models, requiring the engagement of end users for data annotation, which is made simpler by weak supervision. Previous studies have demonstrated the effectiveness of weak labels in improving performance in both disaggregation [24] and multilabel appliance classification tasks [7], [13], and this study proposes using weak labels to jointly distil knowledge and reduce network complexity during transfer learning. The method uses a Convolutional Recurrent Neural Network (CRNN), which has been successfully used in a centralized NILM scenario [7].…”

“…Previous studies have demonstrated the effectiveness of weak labels in improving performance in both disaggregation [24] and multilabel appliance classification tasks [7], [13], and this study proposes using weak labels to jointly distil knowledge and reduce network complexity during transfer learning. The method uses a Convolutional Recurrent Neural Network (CRNN), which has been successfully used in a centralized NILM scenario [7]. In the experiments, the study presents several networks with reduced complexity that retain the main components of the initial model and investigates the trade-off between accuracy and complexity.…”