Our doctoral student Akylas Stratigakos wins the 1st Think SmartGrids 2023 thesis prize

The male prize was awarded to Akylas Stratigakos for his thesis entitled “Towards the Paradigm of Prescriptive Analysis for Energy Forecasting and Power System Optimization”.

Akylas defended his thesis in July 2023 under the supervision of Georges Kariniotakis, Head of Renewable Energies & Smartgrids Group and Andrea Michiorri, Associate Professor at the PERSEE Center of Mines Paris – PSL. The awards were presented in Paris at the Association’s Vœux 2024 ceremony.

Through this award, Think SmartGrids underlines the importance of network digitization, and recognizes the work of the ERSEI group in this field. The group’s research is aimed at facilitating the growing integration of renewable energies into the electricity mix, and simplifying network operation through optimized arbitration based on machine learning and operational research tools.

Congratulations to Akylas Stratigakos for his work!

Summary of the thesis

To mitigate the adverse effects of climate change, the power sector is rapidly moving towards decarbonization through the integration of renewable energy sources, such as wind and solar. In this context, advanced data-driven methods, leveraging the tools of machine learning and operations research, hold great promise as key enablers to address the uncertainty and variability of weather-dependent renewable energy sources. In this thesis, we take a holistic approach by examining the chain of models from data to uncertainty modeling to decisions, and develop data-driven methods that enable improved and resilient decision-making in modern power systems. To maximize the value of forecasts, we develop a method that integrates forecasting and optimization, and propose a framework for assessing the impact of data on decisions. To foster the adoption of advanced data-driven methods and accelerate traditional workflows, we develop an interpretable method for predicting solutions to constrained optimization problems. To strengthen the resilience of models in the face of problematic data, we propose an approach that enables missing data to be handled within an operational framework. We also propose an optimization-based method for clustering data across a number of independent problems, thereby improving overall performance and decision robustness. The proposed methods are validated in various experiments related to power system operation and participation in electricity markets.

Thesis manuscript

Articles from the thesis :

• Akylas Stratigakos, Simon Camal, Andrea Michiorri, Georges Kariniotakis. Prescriptive Trees for Integrated Forecasting and Optimization Applied in Trading of Renewable Energy. IEEE Transactions on Power Systems, 2022, ⟨1109/TPWRS.2022.3152667⟩. https://hal.science/hal-03330017
• Akylas Stratigakos, Panagiotis Andrianesis, Andrea Michiorri, Georges Kariniotakis. Towards Resilient Energy Forecasting: A Robust Optimization Approach. IEEE Transactions on Smart Grid, 2023, pp.1-1. ⟨1109/TSG.2023.3272379⟩. https://hal.science/hal-03792191
• Akylas Stratigakos, Salvador Pineda, Juan Miguel Morales, Georges Kariniotakis. Interpretable Machine Learning for DC Optimal Power Flow with Feasibility Guarantees. IEEE Transactions on Power Systems, 2023, pp.1-12. ⟨1109/TPWRS.2023.3333165⟩. https://hal.science/hal-04038380
• Decision-Focused Data Pooling for Contextual Stochastic Optimization, European Journal of Operational Research https://hal.science/hal-04268454 , (under review)