Abstract:Non-intrusive load monitoring (NILM) is the process of obtaining appliance-level data from a single metering point, measuring total electricity consumption of a household or a business. Appliance-level data can be directly used for demand response applications and energy management systems as well as for awareness raising and motivation for improvements in energy efficiency. Recently, classical machine learning and deep learning (DL) techniques became very popular and proved as highly effective for NILM classification, but with the growing complexity these methods are faced with significant computational and energy demands during both their training and operation. In this paper, we introduce a novel DL model aimed at enhanced multi-label classification of NILM with improved computation and energy efficiency. We also propose an evaluation methodology for comparison of different models using data synthesized from the measurement datasets so as to better represent real-world scenarios. Compared to the state-of-the-art, the proposed model has its energy consumption reduced by more than 23% while providing on average approximately 8 percentage points in performance improvement when evaluating on data derived from REFIT and UK-DALE datasets. We also show a 12 percentage point performance advantage of the proposed DL based model over a random forest model and observe performance degradation with the increase of the number of devices in the household, namely with each additional 5 devices, the average performance degrades by approximately 7 percentage points.

What problem does this paper attempt to address?

The paper primarily focuses on addressing the multi-label classification problem in Non-Intrusive Load Monitoring (NILM), specifically for the on/off state classification of household appliances. Specifically, the research objectives include: 1. **Improving Energy Efficiency**: Develop a novel deep learning model aimed at enhancing computational and energy efficiency in NILM multi-label classification tasks. The new model, compared to existing technologies, significantly reduces energy consumption while ensuring performance improvement. 2. **Enhancing Model Performance**: The proposed deep learning architecture (based on Convolutional Transpose Recurrent Neural Network, CtRNN) was evaluated on the REFIT and UK-DALE datasets. It showed an average performance improvement of approximately 8 percentage points over the state-of-the-art Convolutional Neural Network (CNN) and Convolutional Recurrent Neural Network (CRNN) models, with energy consumption reduced by over 23%. 3. **Extending Evaluation Methods**: A new evaluation method is proposed to generate datasets that are closer to real-world scenarios, better assessing the performance of different models. The new method synthesizes data from measurement datasets, including two evaluation groups—"Realistic Evaluation" (RE) and "Sensitivity Evaluation" (SE). The former considers different numbers of devices, while the latter evaluates the impact of different numbers of active devices on model performance. 4. **Considering Real-World Complexity**: Existing evaluation methods typically focus on 5 common household appliances, whereas the evaluation method proposed in this paper considers a more diverse combination of devices to better reflect actual application scenarios. For example, in some cases, there may be up to 54 different appliances in a household. 5. **Environmentally Friendly Solution**: Given the challenges posed by climate change, the paper emphasizes the importance of reducing energy consumption and proposes a new model that is both effective and energy-efficient, helping to lower overall energy consumption and thus mitigate environmental impact. In summary, this paper aims to address the multi-label classification problem in the NILM field by proposing a novel deep learning model and improved evaluation methods. It not only enhances classification accuracy but also significantly reduces the demand for computational resources, contributing to sustainable development.

Energy Efficient Deep Multi-Label ON/OFF Classification of Low Frequency Metered Home Appliances

Energy Efficient Deep Multi-Label ON/OFF Classification of Low Frequency Metered Home Appliances

Nonintrusive Load Monitoring (NILM) Using a Deep Learning Model with a Transformer-Based Attention Mechanism and Temporal Pooling

Multi-State Energy Classifier to Evaluate the Performance of the NILM Algorithm

A Semi-Supervised Approach for Improving Generalization in Non-Intrusive Load Monitoring

Non-intrusive energy estimation using random forest based multi-label classification and integer linear programming

An active learning framework for the low-frequency Non-Intrusive Load Monitoring problem

Resilient data-driven non-intrusive load monitoring for efficient energy management using machine learning techniques

A Scalable Real-Time Non-Intrusive Load Monitoring System for the Estimation of Household Appliance Power Consumption

Neural Network for NILM Based on Operational State Change Classification

Non-Intrusive Load Monitoring Using a CNN-LSTM-RF Model Considering Label Correlation and Class-Imbalance

Residential Electricity Consumption Behavior Analysis and Energy Management Based on NILM

Deep Learning Based Energy Disaggregation and On/Off Detection of Household Appliances

Low-Complexity Non-Intrusive Load Monitoring Using Unsupervised Learning and Generalized Appliance Models

A weakly supervised active learning framework for non-intrusive load monitoring

Load Identification Based on Attention Semisupervised Curriculum Label Learning With AVME-HT Feature

Non-Intrusive Load Monitoring via Deep Learning Based User Model and Appliance Group Model

MATNilm: Multi-Appliance-Task Non-Intrusive Load Monitoring With Limited Labeled Data

Disaggregating Household Loads Via Semi-Supervised Multi-Label Classification

More Behind Your Electricity Bill: a Dual-DNN Approach to Non-Intrusive Load Monitoring

Non-Intrusive Load Monitoring for Residential Appliances with Ultra-Sparse Sample and Real-Time Computation