The ECLIPSE Digital project recently marked a significant scientific milestone at the 18th International Conference on Electronics, Computers and Artificial Intelligence (ECAI 2026). Held in Bucharest, Romania, from July 2nd to 3rd, 2026, the conference served as a premier platform for experts to discuss the future of smart grids and artificial intelligence.
Representing the project, partners from the University Politehnica of Bucharest (UPB) presented groundbreaking research focused on how to keep consumers engaged in the energy transition without overwhelming them with digital noise.
The Challenge: Notification Fatigue
As part of the Common European Reference Framework (CERF), ECLIPSE DIGITAL aims to empower users through real-time energy awareness. However, a major hurdle in energy-monitoring apps is “notification spam.” Smart meters generate high-frequency data streams that can trigger dozens of alerts daily, leading to behavioral saturation where users eventually ignore messages or even delete the app.
The Solution: The “ANGIE” Intelligent Engine
The paper presented, titled “NN-Assisted Event De-Spamming for CERF-Based Energy-Awareness Notifications: A Proof-of-Concept Implementation,” introduces a sophisticated solution evaluated within the Romanian pilot site.
The research team developed the Advanced Notification Generation Intelligent Engine (ANGIE). This system combines traditional signal filtering (such as EWMA and Butterworth filters) with a lightweight Neural Network (NN) de-spammer layer.
Key findings from the implementation include:
- Drastic Reduction in Noise: In testing, the NN-based method reduced the number of daily alerts from over 111 (raw data) or 30 (traditional filters) down to just 4 to 8 highly relevant notifications.
- Maintaining Relevance: By acting as a “neural de-spammer,” the system filters out redundant alerts while ensuring users still receive critical information at the most impactful moments.
- Improved User Experience: By keeping notifications sparse and meaningful, the framework helps maintain long-term user engagement and prevents “notification fatigue”.
Validation in the Romanian Pilot
These preliminary results, validated through the Romanian pilot, prove the viability of aligning advanced AI management with the CERF standards. This research ensures that the tools we build are not just technically sound, but also psychologically effective for the everyday consumer.
We are proud to see the project consortium contributing to the scientific community and setting the standard for user-centric energy applications in Europe.
📥 Download the Full Paper
Interested in the technical details, the NN architecture, or the filtering formulas? You can access the full preprint version of the research here:
This research was authored by Vlad Sanduleac, Alexandru Raduica, Stefan Popa, Andreea-Georgiana Iantoc, and Mihai Sanduleac (UPB); Dorel Stanescu (ULBS); and Lola Alacreu, Diego Garcia, and Pablo Bort (ETRA).
