Why TSO-DSO Coordination Matters
The growing adoption of decentralized renewable energy sources—like rooftop solar and small-scale wind—has made power flows in distribution grids more complex and less predictable. Effective coordination between TSOs and DSOs is essential to ensure system reliability, efficient energy use, and grid resilience.
Machine learning offers promising tools to manage these dynamics. But using AI in such critical infrastructure also raises high-stakes questions: How do we ensure models remain reliable under uncertainty? Can we generalize to unseen conditions? How do we integrate these tools in a way that’s transparent and accountable?
AI in Action: Research Highlights from MEGAMIND
At the core of Chrysostomou’s research is the development of ML-driven methods that explicitly account for uncertainty and support generalization—two requirements for AI systems operating in the ever-changing conditions of electricity networks.
Key Contributions:
✅ Development of novel algorithms for estimating and utilizing flexibility in active distribution systems.
✅ Implementation of machine learning models such as LSTMs, deep neural networks, and probabilistic neural networks.
✅ Creation of an open-source Python package (TensorConvolutionPlus) to support reproducibility and real-world applicability of his models.
✅ Research on low-observability scenarios, helping DSOs assess system flexibility even with limited data.
✅ Knowledge transfer through workshops on Graph Neural Networks (GNNs) and supervision of MSc research.
Together, these efforts help pave the way for self-managing electricity systems that are both technically sound and socially responsible.
Research in Practice: Open Science and Industry Relevance
A highlight of Chrysostomou’s work is the TensorConvolutionPlus software package, designed to make flexibility area estimation accessible and transparent for grid operators, researchers, and third-party developers. The package is modular and extensible, enabling stakeholders to model, visualize, and plan around grid flexibility—essential for managing congestion, ensuring stability, and maximizing the value of distributed energy resources.
📦 TensorConvolutionPlus – SoftwareX, 2025
This commitment to open science is matched by high-impact publications. His recent paper in IEEE Transactions on Smart Grid presents a method to reliably estimate the feasible combinations of active and reactive power adjustments that DSOs can offer TSOs, even under network constraints.
📄 Aggregated Flexibility Estimation – IEEE Transactions on Smart Grid, 2025
Earlier work presented at IEEE Belgrade PowerTech 2023 explores how limited observability can skew flexibility estimations—an important consideration for real-world deployment.
📄 Operational Flexibility with Low Observability – PowerTech 2023
Looking Ahead: From Algorithms to Action
Chrysostomou’s work illustrates how responsible machine learning can enable a more dynamic, efficient, and equitable energy system—especially when combined with deep domain knowledge and a focus on open, usable tools. As energy systems become increasingly autonomous, his research offers key building blocks for ensuring that AI applications remain robust, interpretable, and aligned with societal values.
Key Takeaways
🔌 AI can empower TSOs and DSOs to make smarter, more responsive decisions in real time.
📊 Uncertainty-aware models are essential for robust grid operation under variable and unforeseen conditions.
🧠 Open-source tools like TensorConvolutionPlus accelerate collaboration and innovation in the energy sector.
🔍 Flexibility estimation must account for both technical constraints and data limitations to be trustworthy.
As the energy transition accelerates, researchers like D. Chrysostomou are helping ensure that the AI behind the grid is not only smart—but also safe, fair, and future-ready.