| _version_ | 1866901391100870656 |
|---|---|
| author | Danish Abdullah |
| author_facet | Danish Abdullah |
| contents | <p><span><span lang="EN-US">Accurate monitoring of energy consumption at the appliance level is essential for sustainable energy management, yet conventional intrusive methods remain costly and impractical for widespread deployment. This study presents a real-time, non-intrusive load monitoring (NILM) framework that integrates artificial intelligence with embedded edge computing to achieve appliance-level energy disaggregation. A sequence-to-point convolutional neural network was implemented on a Raspberry Pi 4B, supported by ESP32-based sensing modules, to process aggregated household energy signals and predict individual appliance usage. The system was trained on a custom dataset collected through an intrusive monitoring stage and subsequently deployed in a real-world residential environment. Evaluation metrics, including Mean Absolute Error (MAE), Normalized Disaggregation Error (NDE), and F1-score, demonstrate reliable prediction accuracy across multiple appliances, with low computational overhead suitable for embedded deployment. The proposed framework addresses privacy concerns by eliminating cloud dependency and reduces implementation costs through edge-based processing. Results highlight the feasibility of deploying NILM systems in smart homes, building automation, and industrial energy management, offering a scalable and sustainable solution for real-time energy analytics.</span></span></p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_17664880 |
| institution | Zenodo |
| language | |
| publishDate | 2025 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Towards Scalable NILM: AI-Integrated Edge Devices for Sustainable Energy Management Danish Abdullah CNN Convolutional neural network Edge computing Energy Disaggregation Non-Intrusive Load Monitoring NILM Raspberry Pi Energy monitoring Realtime Smart Homes Sustainable Energy Management <p><span><span lang="EN-US">Accurate monitoring of energy consumption at the appliance level is essential for sustainable energy management, yet conventional intrusive methods remain costly and impractical for widespread deployment. This study presents a real-time, non-intrusive load monitoring (NILM) framework that integrates artificial intelligence with embedded edge computing to achieve appliance-level energy disaggregation. A sequence-to-point convolutional neural network was implemented on a Raspberry Pi 4B, supported by ESP32-based sensing modules, to process aggregated household energy signals and predict individual appliance usage. The system was trained on a custom dataset collected through an intrusive monitoring stage and subsequently deployed in a real-world residential environment. Evaluation metrics, including Mean Absolute Error (MAE), Normalized Disaggregation Error (NDE), and F1-score, demonstrate reliable prediction accuracy across multiple appliances, with low computational overhead suitable for embedded deployment. The proposed framework addresses privacy concerns by eliminating cloud dependency and reduces implementation costs through edge-based processing. Results highlight the feasibility of deploying NILM systems in smart homes, building automation, and industrial energy management, offering a scalable and sustainable solution for real-time energy analytics.</span></span></p> |
| title | Towards Scalable NILM: AI-Integrated Edge Devices for Sustainable Energy Management |
| topic | CNN Convolutional neural network Edge computing Energy Disaggregation Non-Intrusive Load Monitoring NILM Raspberry Pi Energy monitoring Realtime Smart Homes Sustainable Energy Management |
| url | https://doi.org/10.5281/zenodo.17664880 |