Data-driven uncertainty modelling and control for power system distribution networks
Edgar Salazar Duque defended his PhD thesis at the Department of Electrical Engineering on November 26th.
![[Translate to English:]](https://assets.w3.tue.nl/w/fileadmin/_processed_/4/7/csm_Salazar%20Duque%20Banner%20image_23cce61220.jpg "[Translate to English:]")
The digital transformation of power system distribution networks has allowed distribution system operators to gather an enormous amount of data from advanced metering systems. This data opens new possibilities to address one of the biggest challenges in modern energy systems: uncertainty鈥攖he inherent variability in energy generation and consumption. The PhD thesis of Edgar Salazar Duque explored new ways through which data can be analyzed, visualized and used to create advanced models. These steps help to account for uncertainties, such as changes in solar power due to cloud cover or shifts in electricity demand.
Combining the millions of realistic scenarios that can be created from these models and using the computationally efficient Tensor Power Flow tool, Edgar Salazar Duque and his collaborators were able to create high-quality simulations that are crucial for training reinforcement learning agents, enabling them to make autonomous decisions.
By simulating a wide range of possible future scenarios, the agents learn to adapt to the varying consumption and generation patterns across different areas of the distribution grid, optimizing the control of energy storage systems and ensuring the grid remains resilient and efficient.
The methods developed by Salazar Duque can extend beyond storage control, with applications in demand response strategies, renewable energy forecasting, and grid planning.
Combining data-driven uncertainty modelling, the innovative Tensor Power Flow, and AI techniques, Salazar Duque鈥檚 research paves the way for more resilient, flexible, and sustainable power distribution networks.
Title of PhD thesis: . Supervisors: Han Slootweg, H. Phuong Nguyen, and Anne van der Molen.