EDGE-AEMP: Explainable Edge AI for Low-Carbon Critical Infrastructure

EDGE-AEMP will develop and validate a secure, explainable edge-AI layer that jointly optimises energy consumption and failure risk across heterogeneous mission-critical assets in telecom, data centres, industrial facilities and smart buildings. Operational data is currently fragmented across SmartMeters(OSOS), BMS, SCADA, batteries, UPS, generators, HVAC and IoT devices, while cybersecurity and data-sovereignty constraints often prevent cloud centralisation. The project will create hardware-independent data connectors, local and air-gapped model lifecycle management, multimodal anomaly and remaining-life prediction, explainable root-cause analysis, and energy-optimisation recommendations. It will build on Atlantis AEMP, which has operated in Türk Telekom's production environment for approximately 18 months, monitoring nearly 50,000 devices and processing around 500,000 telemetry records per day. The proposed work goes beyond the existing product through new R&D on cross-site or federated learning with limited labelled failures, edge deployment on heterogeneous hardware, secure model updates, and validation of measurable energy, reliability and maintenance benefits in partner pilot sites. The intended outcome is a commercialisable, interoperable platform validated in at least two countries.