[Paper Review] Adaptive Event Detection for Representative Load Signature Extraction

Lei Yan, Wei Tian+2 more|arXiv (Cornell University)|Jul 23, 2021

Smart Grid Energy Management25 references4 citations

TL;DR

This paper proposes an adaptive event detection method, WAMMA, for non-intrusive load monitoring (NILM) that dynamically adjusts window width, margin width, and threshold to accurately detect transient events in high-rate residential load data (≥1Hz). The method outperforms state-of-the-art approaches on 20Hz, 50Hz, and 60Hz datasets, enabling robust extraction and quantification of representative transient and steady-state load signatures for improved NILM accuracy and load reconstruction.

ABSTRACT

Event detection is the first step in event-based non-intrusive load monitoring (NILM) and it can provide useful transient information to identify appliances. However, existing event detection methods with fixed parameters may fail in case of unpredictable and complicated residential load changes such as high fluctuation, long transition, and near simultaneity. This paper proposes a dynamic time-window approach to deal with these highly complex load variations. Specifically, a window with adaptive margins, multi-timescale window screening, and adaptive threshold (WAMMA) method is proposed to detect events in aggregated home appliance load data with high sampling rate (>1Hz). The proposed method accurately captures the transient process by adaptively tuning parameters including window width, margin width, and change threshold. Furthermore, representative transient and steady-state load signatures are extracted and, for the first time, quantified from transient and steady periods segmented by detected events. Case studies on a 20Hz dataset, the 50Hz LIFTED dataset, and the 60Hz BLUED dataset show that the proposed method can robustly outperform other state-of-art event detection methods. This paper also shows that the extracted load signatures can improve NILM accuracy and help develop other applications such as load reconstruction to generate realistic load data for NILM research.

Motivation & Objective

• To address the limitations of fixed-parameter event detection in handling complex residential load variations such as high fluctuation, long transitions, and near-simultaneous events.
• To develop a dynamic time-window approach that adapts to varying load dynamics in real-time.
• To extract and quantitatively characterize representative transient and steady-state load signatures from event-segmented data for enhanced NILM performance.
• To demonstrate the utility of extracted signatures in improving NILM accuracy and supporting load reconstruction applications.

Proposed method

• Proposes a window with adaptive margins, multi-timescale window screening, and adaptive threshold (WAMMA) method for event detection in high-sampling-rate load data.
• Uses a dynamic window width that adjusts based on signal characteristics to capture transient events without missing or overfitting.
• Incorporates adaptive margins and thresholds that vary with local signal energy and change rate to enhance detection robustness.
• Applies multi-timescale analysis to detect both fast and slow transitions by comparing signal changes across different temporal scales.
• Segments load data into transient and steady-state periods based on detected events to extract representative signatures.
• Quantifies extracted signatures from both transient and steady-state phases for use in downstream NILM and load reconstruction tasks.

Experimental results

Research questions

• RQ1 Can an adaptive event detection method outperform fixed-parameter methods in detecting transient events under high load variability?
• RQ2 How effectively can the proposed WAMMA method handle complex residential load dynamics such as long transition periods and near-simultaneous appliance operations?
• RQ3 To what extent do the extracted transient and steady-state load signatures improve NILM accuracy compared to conventional methods?
• RQ4 Can the extracted load signatures be used to generate realistic synthetic load data for NILM research and applications?

Key findings

• The WAMMA method achieves superior event detection performance across multiple high-rate datasets, including 20Hz, 50Hz (LIFTED), and 60Hz (BLUED) data, outperforming state-of-the-art methods.
• The method successfully captures transient processes with high accuracy by dynamically tuning window width, margin width, and change threshold.
• Representative load signatures extracted from transient and steady-state phases are quantified for the first time, enabling precise characterization of appliance behavior.
• The use of extracted signatures improves NILM accuracy, demonstrating their value in load disaggregation tasks.
• The method enables realistic load reconstruction by generating synthetic load data based on extracted signatures, supporting future NILM research.