Fast and Robust Security Constrained AC Optimal Power Flow for the Dutch Electricity Grid
Two master thesis projects: one mathematics oriented, one AI oriented.
Introduction
TenneT is a Transmission System Operator (TSO) that designs, builds, operates and manages the high-voltage electricity grid in the Netherlands and large parts of Germany with the fundamental purpose of reliably and efficiently transporting electric power at all times. Due to the energy transition and rapid expansion of the energy sector that are pushing the transmission grid to its limits, the need for optimizing the grid has in recent times been greater than ever. Optimal power flow calculations play an important role in facilitating efficient transport of electric power.
AC Optimal Power Flow
Optimal power flow in the simplest definition involves determining operating setpoints for controllable assets in the grid such as generators, shunts and transformers that lead to minimal operational costs while satisfying demand and operational constraints such as voltages and line loading within safety bounds. As the electricity grid is predominantly based on Alternating Current (AC), the underlying physics of the grid is non- linear making the AC optimal power flow problem non-linear and possibly non-convex. Hence, solving the AC-OPF problem is non-trivial, computationally expensive and prone to divergence. Furthermore, when security constraints are imposed over time, solving the AC-OPF problem can become prohibitively slow. This stimulates the use of approximate methods that compromise on accuracy for speed.
Master thesis projects
We are looking for two master students to work on SC-ACOPF problems for the Dutch transmission grid along the following two tracks, one student a track each.
- Track Mathematical Optimization. This track involves developing classical optimization techniques to speed up SC-ACOPF calculations. The student will be building on the work of a previous master student who designed a solid & flexible framework for ACOPF calculations (without security constraints) for the Dutch grid. In this project, security constraints will be added to the existing framework and the calculations sped up.
- Track Graph Neural Networks. This track involves training graph neural networks to speed up SC-ACOPF calculations with reasonable accuracy. The student will be working on an internally developed module that trains GNNs for performing AC power flow calculations. In this project, the student will be extending the module to ACOPF with security constraints.
The objective is to design a calculation framework each that is fast and robust enough to be used close to real-time grid operation and also investigate convergence problems. Working on two tracks in parallel has the advantage of comparing classical and machine learning approaches for OPF from the start, eventually leading to a solid assessment of their efficacy for optimizing the grid. This is valuable for grid operators to choose between classical and machine learning algorithms for critical applications.
The minimum duration for the master thesis projects is six months. (Nine months is preferred).
Approach:
1. Literature review on AC optimal power flow problems. 2. Solving SC-ACOPF problems for small test networks. 3. Solving SC-ACOPF for the onshore Dutch TenneT grid. 4. Speeding up calculations for the Dutch grid and solving convergence problems.
Requirements:
1. Background in computing sciences, mathematics, machine learning, physics or engineering. 2. Good programming skills in Python, knowledge of libraries for deep learning is a plus. 3. Knowledge on power flow analysis or network calculations is helpful, but not required. 4. Interest in the energy transition is a big plus.
Deliverables: Each student is expected to deliver the following: 1. Literature review on AC optimal power flow problems. 2. Python code with implementations for solving SC-ACOPF problems. 3. MSc thesis describing the results. 4. Presentation of the results at TenneT office in Arnhem.
Workplace: The project is to be performed in the ODINA team. Daily supervision is by Shravan Chipli. ODINA meetings are to be attended twice a week (Tuesdays and Thursdays). Weekly supervision by the ODINA team during live meetings at Arnhem office is preferred, at least once a week. Daily contact and supervision is live at Arnhem office or by web-meetings.
Contact: Yuliya Shapovalova.