At the Portfolio AI in Energy 2026 conference, E.On data analysts Káplár‑Zagyi Evelin and Székely Magor Illés presented how artificial intelligence can support distribution system operators (DSOs) in optimizing network reconstruction and vegetation management based on data.
What AI can achieve in these areas
According to the presentation, vegetation management primarily aims to reduce operational safety risks, ensure regulatory compliance, and optimize resource allocation. Network reconstruction focuses on efficient CAPEX use, improving service levels, and enabling proactive maintenance.
Digitalization as the foundation
The speakers emphasized that digitalization is the essential prerequisite for data‑driven decision support. Several international initiatives address this, including the Danube InGrid Project (Danube Intelligent Grid), a cross‑border Hungarian–Slovak smart grid collaboration in which E.On participates.
Key building blocks include:
- smart network elements (IoT) providing real‑time and recorded measurements;
- a 3D digital twin of the network, which goes beyond replacing paper records to represent the entire system digitally.
Data and AI governance
The next step is establishing data and AI governance: regulatory and technological frameworks, a common language, risk management and compliance. AI decision support is implemented under human oversight.
The AI ecosystem, as described by the presenters, comprises layers such as deep learning, machine learning, general AI components and AI governance. Data required for vegetation management, image processing and network reconstruction are the subject of machine learning modelling.
Data quality and structure
Important aspects of data governance are:
- data structure and inventory;
- data quality checks and consistency;
- handling data volume.
A frequent technical challenge is siloed and “dirty” data; a central data warehouse can address this. The changing corporate environment — new systems and new data structures — can be handled by implementing data governance.
Expert oversight, transparency and scalability
To address engineers’ skepticism and the “Black Box” concern, regular expert consultations, incorporation of expert opinions, testing and back‑measurements are recommended. Cloud‑based architectures can meet scalability and compute capacity requirements.
Time‑series snapshot approach and predictive modelling
The presenters explained the time‑series snapshot logic: a dynamic data model records snapshots of vegetation and network state at given points in time. Predictive modelling is based on reconstructing historical states. The principle of time‑correctness means models only use information available up to the snapshot time.
Using multiple snapshots produces a more stable training set and improves model training: identifying vegetation growth trends and changes in network state becomes more accurate when multiple time points are included.
Conclusion
Káplár‑Zagyi Evelin and Székely Magor Illés concluded that applying AI and structured data management at DSOs can increase safety, enable cost‑efficient investments and improve operations. Success depends on digitalization, high‑quality data, governance frameworks, expert oversight and scalable computing infrastructure.
