Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fusion-ResNet: A Lightweight multi-label NILM Model Using PCA-ICA Feature Fusion

Sahar Moghimian Hoosh, Ilia Kamyshev, Henni Ouerdane

TL;DR

The paper tackles the NILM challenge of accurately disaggregating high-frequency household energy use, especially under many simultaneous appliances, by introducing a PCA-ICA feature fusion (NILM-ICPC) and a lightweight residual network (Fusion-ResNet). The approach combines PCA and ICA features, ranked by kurtosis, into a compact input for a residual feed-forward classifier with 18 blocks and ~65k parameters, optimizing with Binary Cross-Entropy and a 0.5 threshold for multi-label appliance detection. Empirical results on high-frequency PLAID data show that Fusion-ResNet with NILM-ICPC outperforms baselines (FIT-PS+LSTM, Fryze+CNN) with an average $F_1$ around $0.77$, while exhibiting lower training and inference costs and robustness to up to 15 concurrent devices. The findings suggest a practical, edge-friendly NILM solution that leverages physics-aware feature fusion to improve generalization and efficiency for real-world energy disaggregation scenarios.

Abstract

Non-intrusive load monitoring (NILM) is an advanced load monitoring technique that uses data-driven algorithms to disaggregate the total power consumption of a household into the consumption of individual appliances. However, real-world NILM deployment still faces major challenges, including overfitting, low model generalization, and disaggregating a large number of appliances operating at the same time. To address these challenges, this work proposes an end-to-end framework for the NILM classification task, which consists of high-frequency labeled data, a feature extraction method, and a lightweight neural network. Within this framework, we introduce a novel feature extraction method that fuses Independent Component Analysis (ICA) and Principal Component Analysis (PCA) features. Moreover, we propose a lightweight architecture for multi-label NILM classification (Fusion-ResNet). The proposed feature-based model achieves a higher $F1$ score on average and across different appliances compared to state-of-the-art NILM classifiers while minimizing the training and inference time. Finally, we assessed the performance of our model against baselines with a varying number of simultaneously active devices. Results demonstrate that Fusion-ResNet is relatively robust to stress conditions with up to 15 concurrently active appliances.

Fusion-ResNet: A Lightweight multi-label NILM Model Using PCA-ICA Feature Fusion

TL;DR

The paper tackles the NILM challenge of accurately disaggregating high-frequency household energy use, especially under many simultaneous appliances, by introducing a PCA-ICA feature fusion (NILM-ICPC) and a lightweight residual network (Fusion-ResNet). The approach combines PCA and ICA features, ranked by kurtosis, into a compact input for a residual feed-forward classifier with 18 blocks and ~65k parameters, optimizing with Binary Cross-Entropy and a 0.5 threshold for multi-label appliance detection. Empirical results on high-frequency PLAID data show that Fusion-ResNet with NILM-ICPC outperforms baselines (FIT-PS+LSTM, Fryze+CNN) with an average around , while exhibiting lower training and inference costs and robustness to up to 15 concurrent devices. The findings suggest a practical, edge-friendly NILM solution that leverages physics-aware feature fusion to improve generalization and efficiency for real-world energy disaggregation scenarios.

Abstract

Non-intrusive load monitoring (NILM) is an advanced load monitoring technique that uses data-driven algorithms to disaggregate the total power consumption of a household into the consumption of individual appliances. However, real-world NILM deployment still faces major challenges, including overfitting, low model generalization, and disaggregating a large number of appliances operating at the same time. To address these challenges, this work proposes an end-to-end framework for the NILM classification task, which consists of high-frequency labeled data, a feature extraction method, and a lightweight neural network. Within this framework, we introduce a novel feature extraction method that fuses Independent Component Analysis (ICA) and Principal Component Analysis (PCA) features. Moreover, we propose a lightweight architecture for multi-label NILM classification (Fusion-ResNet). The proposed feature-based model achieves a higher score on average and across different appliances compared to state-of-the-art NILM classifiers while minimizing the training and inference time. Finally, we assessed the performance of our model against baselines with a varying number of simultaneously active devices. Results demonstrate that Fusion-ResNet is relatively robust to stress conditions with up to 15 concurrently active appliances.

Paper Structure

This paper contains 19 sections, 25 equations, 8 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Datasets
  4. Feature extraction
  5. Principal Component Analysis (PCA)
  6. Independent Component Analysis (ICA)
  7. Fusion of PCA and ICA Features (NILM-ICPC)
  8. Algorithms and Models
  9. Proposed Model: Fusion-ResNet
  10. Training objective
  11. Inference
  12. Optimization and training
  13. Evaluation
  14. Baseline models
  15. FIT-PS+LSTM
  16. ...and 4 more sections

Figures (8)

  • Figure 1: Architecture of the proposed model, Fusion-ResNet. The number of parameters for this model is 65,000.
  • Figure 2: BCE loss for the training set.
  • Figure 3: BCE loss for the validation set
  • Figure 4: $F_1$ score (sample average) for train set.
  • Figure 5: $F_1$ score (sample average) for validation set.
  • ...and 3 more figures