[Paper Review] A Survey on Non-Intrusive Load Monitoring Methodies and Techniques for Energy Disaggregation Problem

Anthony Faustine, Nerey H. Mvungi+2 more|arXiv (Cornell University)|Mar 2, 2017

Smart Grid Energy Management50 references121 citations

TL;DR

This paper surveys NILM systems, methods, metrics, benchmarking tools, datasets, and future directions for energy disaggregation.

ABSTRACT

The rapid urbanization of developing countries coupled with explosion in construction of high rising buildings and the high power usage in them calls for conservation and efficient energy program. Such a program require monitoring of end-use appliances energy consumption in real-time. The worldwide recent adoption of smart-meter in smart-grid, has led to the rise of Non-Intrusive Load Monitoring (NILM); which enables estimation of appliance-specific power consumption from building's aggregate power consumption reading. NILM provides households with cost-effective real-time monitoring of end-use appliances to help them understand their consumption pattern and become part and parcel of energy conservation strategy. This paper presents an up to date overview of NILM system and its associated methods and techniques for energy disaggregation problem. This is followed by the review of the state-of-the art NILM algorithms. Furthermore, we review several performance metrics used by NILM researcher to evaluate NILM algorithms and discuss existing benchmarking framework for direct comparison of the state of the art NILM algorithms. Finally, the paper discuss potential NILM use-cases, presents an overview of the public available dataset and highlight challenges and future research directions.

Motivation & Objective

• Motivate real-time, non-intrusive end-use energy monitoring in buildings to support conservation and policy evaluation.
• Provide an up-to-date overview of NILM architectures, methods, and techniques for energy disaggregation.
• Review state-of-the-art NILM algorithms across categories (HMM-based, graph signal processing, and deep learning).
• Discuss evaluation metrics and benchmarking frameworks to enable fair comparison of NILM methods.
• Highlight public datasets, use-cases, challenges, and avenues for future research.

Proposed method

• Classify NILM approaches into event-based and state-based event detectors.
• Describe appliance signatures and the distinction between transient and steady-state features.
• Survey learning/inference paradigms (supervised vs unsupervised; HMM, FHMM, and deep learning hybrids).
• Discuss graph signal processing (GSP) applications in NILM and their limitations.
• Summarize deep learning architectures (RNNs, CNNs, autoencoders, and hybrid HMM/DNN models).
• Outline evaluation metrics and the importance of benchmarking tools like NILMTK and NILM-Eval.

Experimental results

Research questions

• RQ1 What are the main NILM approaches used to disaggregate energy from aggregate power readings?
• RQ2 How are NILM models learned and inferred, and what are the trade-offs between supervised and unsupervised methods?
• RQ3 What metrics and benchmarking frameworks exist to fairly evaluate NILM algorithms?
• RQ4 What public datasets are available for NILM research, and what use-cases do they support?
• RQ5 What challenges and future directions shape NILM research and deployment in real-world buildings?

Key findings

• HMM-based and FHMM approaches dominate unsupervised NILM, with extensions and variants to address multi-state appliances.
• Graph Signal Processing and deep learning offer alternative NILM paradigms, each with advantages and limitations in training data and real-time inference.
• A wide range of evaluation metrics exist (accuracy, F-measure, RMSE, de, EEFI) and there is a lack of standardized benchmarking.
• Open-source benchmarking toolkits NILMTK and NILM-Eval enable reproducible comparisons but have limitations and integration gaps.
• A variety of public datasets (REDD, UK-DALE, UK-DALE, REFIT, GREEND, AMPDS, etc.) support diverse sensing resolutions and contexts.
• The review highlights challenges such as scalability to many appliances, need for labeled data, noise sensitivity, and real-time applicability, outlining future research directions.