VSL develops test and validation framework for reliable data analysis methods in electricity grids

Publication

Artificial intelligence (AI) and data analytics are becoming increasingly important in the operation of electricity grids. To ensure that these applications remain reliable and fit for purpose, VSL is developing a uniform test and validation framework for grid data analytics together with European partners. The first results have been published in a paper accepted for the CIRED 2026 Brussels Workshop.

The framework is being developed within the European research project Metrology for Reliable Grid Data Analytics (GridData) and focuses on testing and validating data-driven methods for critical applications, such as congestion forecasting in distribution networks. Key aspects include uncertainty quantification, input data quality, model robustness and the use of explainable AI techniques.

The research shows that traditional one-off software verification is insufficient for AI and machine learning models operating in dynamic, high-dimensional environments. The proposed framework combines best practices from metrology and machine learning to support a structured and ongoing assessment of model reliability.

With this work, VSL contributes to the development of trustworthy AI solutions for critical infrastructure and supports future regulatory requirements, including those anticipated under the European AI Act.Met dit werk levert VSL een belangrijke bijdrage aan de ontwikkeling van betrouwbare AI-toepassingen voor kritische infrastructuur en speelt het in op toekomstige eisen vanuit onder andere de Europese AI Act.

“This paper is a pre-print of a paper submitted to and accepted for publication in the Conference Proceedings of the CIRED 2026 Brussels Workshop on Implementing Successful Innovation in Distribution Networks and is subject to the Organizers’ Copyright. The copy of record will be available at The IET Digital Library.”