MSc thesis project proposal

Network anomaly localisation with application to smart grids

Anomaly (and, especially, outage) detection is considered one of the cornerstone tasks in tomorrow’s smart networks. Unexpected events (tree falling, breaker failures) can make lines or part of the grid inoperative and cause cascade breakdowns within a few minutes. Developing methods able to promptly detect failures in smart networks is, therefore, a key task. Current works find line failures through phasor angle measurements and usually use exhaustive search methods to classify the line status. In this project, we will study graph-based detection approaches to detect (a) if a link outage happens, (b) in which area of the grid the failure occurs. With the aim of reducing the sensing costs (i.e., place fewer meters in the grid), facilitating concurrent and distributed detectors, and making the algorithms robust to missing data, we will address these tasks when only a few measurements are present. An efficient solution to this task is timely and aligns well with the European strategy for smart energy management. As such is highly appreciated in both academic and industry works since it is lined with practical challenges. Moreover, you will gain expertise in tools such as smart grids, network science, sparse representation, sparse sensing, and detection theory. This knowledge will enrich your background gained during the master classes and prepare you for the job market.


The problem formulation for locating anomalies in networks is not difficult to formulate, while a few challenges are present when a few measurements should be used. In this project, your task would be to derive an accurate model for locating anomalies in a network (especially in smart grids) based on dictionary representation and graph signal processing. You will work with both simulated and real data and you are requested to compare the performance with other state-of-the-art approaches. More in particular, you will adopt the following approach: (a) Test if a graph-tight dictionary leads to a sparse representation of the real power injections. (b) Exploit such dictionary representation for detecting anomalies in each vertex of the network. (c) Study the tradeoff between classification accuracy, network topology, and sparsity. (d) Develop robust detectors where few measurements can be collected. (e) Evaluate these detectors through the toolbox MATPOWER on two benchmark systems: the IEEE 118-bus and the PTSCA 300-bus system.


To be able to successfully complete such task, you should fulfill the following requirements: - be a self-motivated student with a desire in making an impactful master thesis; - have strong linear algebra background; - have a good knowledge of tools applied from statistical signal processing, estimation and detection theory, and convex optimization; - respect deadlines in delivering the results; - advanced coding skills in Matlab, Python (preferred), R, on any other programming language.

Contact Geert Leus

Signal Processing Systems Group

Department of Microelectronics

Last modified: 2023-01-13