Context
Low‑voltage distribution networks increasingly host flexible resources that can support grid operation, but their effective use requires knowledge of the actual network state.
Challenge
Using flexibility installed in low‑voltage networks requires knowledge of the actual network state. Traditional measurement concepts are not cost‑efficient in low‑voltage grids.
AI Solution
Artificial neural networks estimate the state of a low‑voltage network using only one real‑time measurement in secondary substations for inference. The model is trained using exogenous inputs such as temperature and solar irradiance measurements and historical consumption and generation profiles; smart‑meter information can be used to simplify training.
Impact & evidence
The solution contributes to grid stability and system flexibility and enables faster operations through automation in low‑voltage network management. Impact is evidenced by state‑estimation accuracy measured through root mean square error (RMSE) and maximum component error (Maxmax E), with proof‑of‑concept results on real grid test data showing RMSE below 0.2% and Maxmax E of approximately 1.5%.
Implementation / Adoption
Implementation requires a network information or management system with appropriate interfaces and a consistent network data model or digital twin of the low‑voltage distribution grid. Smart‑meter data can be used to support and simplify the training of artificial neural networks, while adoption within the IT/OT environment of a distribution system operator depends on the interfaces provided by existing systems to enable integration and operational deployment of the solution.
Data
Solution requires relevant data models / digital twin of the low voltage networks and/or smart meter information
