# Energy Injection Identification enabled Disaggregation with Deep Multi-Task Learning

Source: [Energy Injection Identification enabled Disaggregation with Deep Multi-Task Learning](<https://www.sota2.com/research/paper/68fa7ba71e209131cea89aa7>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-24T05:00:22.883Z

Authors: Xudong Wang, Guoming Tang, Junyu Xue, Srinivasan Keshav, Tongxin Li, Chris Ding

Published: 2025-08-20T10:35:38.000Z

## Abstract

Non-Intrusive Load Monitoring (NILM) offers a cost-effective method to obtain fine-grained appliance-level energy consumption in smart homes and building applications. However, the increasing adoption of behind-the-meter (BTM) energy sources such as solar panels and battery storage poses new challenges for conventional NILM methods that rely solely on at-the-meter data. The energy injected from the BTM sources can obscure the power signatures of individual appliances, leading to a significant decrease in NILM performance. To address this challenge, we present DualNILM, a deep multi-task learning framework designed for the dual tasks of appliance state recognition and injected energy identification. Using a Transformer-based architecture that integrates sequence-to-point and sequence-to-sequence strategies, DualNILM effectively captures multiscale temporal dependencies in the aggregate power consumption patterns, allowing for accurate appliance state recognition and energy injection identification. Extensive evaluation on self-collected and synthesized datasets demonstrates that DualNILM maintains an excellent performance for dual tasks in NILM, much outperforming conventional methods. Our work underscores the framework's potential for robust energy disaggregation in modern energy systems with renewable penetration. Synthetic photovoltaic augmented datasets with realistic injection simulation methodology are open-sourced at https://github.com/MathAdventurer/PV-Augmented-NILM-Datasets.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Appliance State Recognition on REDD House 2 Synthetic Injection](<https://www.sota2.com/research/sota/appliance-state-recognition-on-redd-house-2-synthetic-injection.md>): Appliance State Recognition; REDD House 2 Synthetic Injection; Accuracy: 99.7.

- [Energy Injection Disaggregation on REDD Synthetic Injection (House 1)](<https://www.sota2.com/research/sota/energy-injection-disaggregation-on-redd-synthetic-injection-house-1.md>): Energy Injection Disaggregation; REDD Synthetic Injection (House 1); Avg RMSE: 0.6661.

- [Energy Injection Disaggregation on REDD Synthetic Injection (House 2)](<https://www.sota2.com/research/sota/energy-injection-disaggregation-on-redd-synthetic-injection-house-2.md>): Energy Injection Disaggregation; REDD Synthetic Injection (House 2); Average RMSE: 0.7065.

- [Energy Injection Disaggregation on REDD Synthetic Injection (House 3)](<https://www.sota2.com/research/sota/energy-injection-disaggregation-on-redd-synthetic-injection-house-3.md>): Energy Injection Disaggregation; REDD Synthetic Injection (House 3); Avg. RMSE: 0.4261.

- [Energy Injection Disaggregation on UKDALE Synthetic Injection (House 1)](<https://www.sota2.com/research/sota/energy-injection-disaggregation-on-ukdale-synthetic-injection-house-1.md>): Energy Injection Disaggregation; UKDALE Synthetic Injection (House 1); Avg. RMSE: 0.9008.

- [Appliance State Recognition on Laboratory Dataset Real Injection](<https://www.sota2.com/research/sota/appliance-state-recognition-on-laboratory-dataset-real-injection.md>): Appliance State Recognition; Laboratory Dataset Real Injection; Accuracy: 98.54.

- [Appliance State Recognition on REDD House 1 Synthetic Injection](<https://www.sota2.com/research/sota/appliance-state-recognition-on-redd-house-1-synthetic-injection.md>): Appliance State Recognition; REDD House 1 Synthetic Injection; Accuracy: 99.84.

- [Appliance State Recognition on REDD House 3 Synthetic Injection](<https://www.sota2.com/research/sota/appliance-state-recognition-on-redd-house-3-synthetic-injection.md>): Appliance State Recognition; REDD House 3 Synthetic Injection; Accuracy: 99.69.

- [Appliance State Recognition on UKDALE House 1 (Synthetic Injection)](<https://www.sota2.com/research/sota/appliance-state-recognition-on-ukdale-house-1-synthetic-injection.md>): Appliance State Recognition; UKDALE House 1 (Synthetic Injection); Accuracy: 99.72.

- [Appliance State Recognition on UKDALE House 2 Synthetic Injection](<https://www.sota2.com/research/sota/appliance-state-recognition-on-ukdale-house-2-synthetic-injection.md>): Appliance State Recognition; UKDALE House 2 Synthetic Injection; Accuracy: 99.73.

- [Energy Injection Identification on Laboratory Dataset Real BTM Injection](<https://www.sota2.com/research/sota/energy-injection-identification-on-laboratory-dataset-real-btm-injection.md>): Energy Injection Identification; Laboratory Dataset Real BTM Injection; Accuracy: 98.54.

- [Energy Injection Identification on REDD Synthetic PV Injection House 1](<https://www.sota2.com/research/sota/energy-injection-identification-on-redd-synthetic-pv-injection-house-1.md>): Energy Injection Identification; REDD Synthetic PV Injection House 1; Accuracy: 99.84.

- [Energy Injection Identification on REDD (Synthetic PV Injection) House 3](<https://www.sota2.com/research/sota/energy-injection-identification-on-redd-synthetic-pv-injection-house-3.md>): Energy Injection Identification; REDD (Synthetic PV Injection) House 3; Accuracy: 99.69.

- [Energy Injection Identification on UKDALE Synthetic PV Injection House 1](<https://www.sota2.com/research/sota/energy-injection-identification-on-ukdale-synthetic-pv-injection-house-1.md>): Energy Injection Identification; UKDALE Synthetic PV Injection House 1; Accuracy: 99.72.

- [Energy Injection Identification on UKDALE Synthetic PV Injection House 2](<https://www.sota2.com/research/sota/energy-injection-identification-on-ukdale-synthetic-pv-injection-house-2.md>): Energy Injection Identification; UKDALE Synthetic PV Injection House 2; Accuracy: 99.73.

- [State Recognition on REDD House 1](<https://www.sota2.com/research/sota/state-recognition-on-redd-house-1.md>): State Recognition; REDD House 1; Accuracy: 99.84.

- [State Recognition on REDD House 2](<https://www.sota2.com/research/sota/state-recognition-on-redd-house-2.md>): State Recognition; REDD House 2; Accuracy: 99.7.

- [State Recognition on REDD House 3](<https://www.sota2.com/research/sota/state-recognition-on-redd-house-3.md>): State Recognition; REDD House 3; Accuracy: 99.69.

- [State Recognition on REDD Overall Avg (all splits)](<https://www.sota2.com/research/sota/state-recognition-on-redd-overall-avg-all-splits.md>): State Recognition; REDD Overall Avg (all splits); Accuracy: 99.74.

- [State Recognition on UKDALE (House 1)](<https://www.sota2.com/research/sota/state-recognition-on-ukdale-house-1.md>): State Recognition; UKDALE (House 1); Accuracy: 99.72.

- [State Recognition on UKDALE (House 2)](<https://www.sota2.com/research/sota/state-recognition-on-ukdale-house-2.md>): State Recognition; UKDALE (House 2); Accuracy: 99.73.

- [State Recognition on UKDALE](<https://www.sota2.com/research/sota/state-recognition-on-ukdale.md>): State Recognition; UKDALE; Accuracy: 99.73.

- [Energy Injection Disaggregation on Laboratory Dataset Real Injection](<https://www.sota2.com/research/sota/energy-injection-disaggregation-on-laboratory-dataset-real-injection.md>): Energy Injection Disaggregation; Laboratory Dataset Real Injection; Avg. RMSE: 0.1429.

- [Injection Disaggregation on REDD Overall Avg (all splits)](<https://www.sota2.com/research/sota/injection-disaggregation-on-redd-overall-avg-all-splits.md>): Injection Disaggregation; REDD Overall Avg (all splits); RMSE: 0.5996.

- [Injection Disaggregation on UKDALE](<https://www.sota2.com/research/sota/injection-disaggregation-on-ukdale.md>): Injection Disaggregation; UKDALE; Average RMSE: 0.9176.

- [Energy Injection Disaggregation on UKDALE Synthetic Injection (House 2)](<https://www.sota2.com/research/sota/energy-injection-disaggregation-on-ukdale-synthetic-injection-house-2.md>): Energy Injection Disaggregation; UKDALE Synthetic Injection (House 2); Average RMSE: 0.9344.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2508.14600>)

[PDF](<https://arxiv.org/pdf/2508.14600.pdf>)
