Submission 359
Secure Smart Meter Infrastructure as Grid Intelligence Layer: NILM-Based Disaggregation and Modular HEMS Integration in a Live German Grid State Monitoring Deployment
71 GIW26-359
Presented by: Andre Hoffmann
Germany's rollout of intelligent metering systems (iMS) establishes a uniquely regulated smart meter infrastructure combining the smart meter gateway (SMGw) — a certified security anchor — with standardized metering profiles such as TAF10, which enables 1-minute resolution grid state measurements (power, current, frequency, phase angles) and bidirectional control via home energy management systems (HEMS). This regulatory architecture creates a cryptographically secured, end-to-end observable data chain from household meter to grid operator backend, yet its potential for active distribution system management remains largely unexplored in international literature.
This paper presents the SECProMo project, a BMWE-funded research initiative conducted by Stadtwerke Bielefeld, SmartOptimo, Hochschule Osnabrück, PPC, Items, and Arvato Systems, deploying 20 iMS gateways with TAF10 1-minute grid state values at a live low-voltage transformer in Bielefeld. Three contributions are made.
First, we elaborate the German iMS security chain in an internationally accessible framework: the SMGw enforces role-based data access across WAN, HAN, and LMN interfaces, with BSI TR-03109 defining cryptographic requirements for all communication paths. TAF10 data flows to grid operator backends — delivering high-resolution grid state data without compromising household privacy.
Second, we present a NILM-based disaggregation framework that estimates device-class composition across approximately 450 connected households from transformer-level aggregates using a hybrid NNLS/XGBoost architecture on P-Q feature vectors. Reactive power signatures prove essential for discriminating controllable loads — heat pumps, EV chargers, and PV inverters — that overlap substantially in active power space. The full ingestion pipeline is operational across all 20 iMS; systematic forecasting validation is ongoing. Grid state simulations based on pandapower enable dynamic local grid fees and power limits (Flexband) to be derived for day-ahead planning.
Third, we present a modular HEMS architecture for bidirectional control behind the SMGw. Commercial operator-grade systems are compared to an open-source stack — Home Assistant with EVCC — as a lean, locally-executed control agent. Three energy management use cases are implemented and investigated: (i) solar surplus charging, (ii) price-driven load optimization using EPEX Spot intraday as well as dynamic grid fee signals, and (iii) §42c EnWG-compliant Energy Sharing across multi-household communities based on calibrated smart meter data without additional sub-metering. Crucially, the co-simulation framework enables systematic evaluation of how these different energy management service use cases — individually and in combination — affect local grid quality, congestion probability, and balancing group efficiency, providing evidence-based guidance for regulatory design and DSO deployment strategies.
Together, SECProMo demonstrates how Germany's regulated iMS-infrastructure can serve as a foundation for both grid observability and decentralized energy market participation — offering a transferable model for jurisdictions designing next-generation grid usage and feed-in billing regulation.