What question did this study set out to answer?

The central aim is to develop a reproducible algorithm that detects ON/OFF states of electrical loads from aggregated active power.

March 7, 2026ElectricityOpen Access

NILM-Based Feedback for Demand Response: A Reproducible Binary State-Detection Algorithm Using Active Power

View Full Paper

Ask AI

Bookmark

The central aim is to develop a reproducible algorithm that detects ON/OFF states of electrical loads from aggregated active power.
Developed a pipeline using hysteresis-based labeling and adaptive thresholds for state detection.
Implemented compact feature engineering focused on global active power.
Utilized permutation importance to select informative predictors after collinearity removal.
Trained several probabilistic binary classifiers, including LGBM and XGBoost, to identify load states.
Applied post-processing techniques to stabilize predictions and reduce false triggers.
Experiments showed that LGBM and other gradient boosting models provide reliable state estimates.
The models effectively balanced missed events and false alarms through optimized decision thresholds.
Performance was evaluated using sample-wise and event-based metrics for accurate switching interval detection.

No takes yet. Share an insight, caveat, or question.

Demonstrates a binary state-detection algorithm for demand response, utilizing active power signals to enhance load monitoring accuracy.

Zhukovskiy et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark