Student Projects

Extreme weather events, such as flooding, heavy snowfall, and storms, increasingly disrupt transportation networks, posing significant threats to regional mobility and economic stability. These disruptions propagate through interconnected systems, with widespread consequences on infrastructure and services. The rise of electric vehicles (EVs) presents new challenges, as charging infrastructure may become a critical bottleneck during emergencies, leading to prolonged delays and reduced evacuation efficiency. Developing resilient transport systems is imperative to minimise these impacts and ensure reliable mobility in the face of climate uncertainty.

This Master’s thesis project aims to build a comprehensive transport system model that captures the complex interactions between EV infrastructure, traffic flow, and driver behaviour under extreme weather conditions. By integrating scenario-based simulations, the model will assess strategies to mitigate congestion, optimise charging infrastructure, and enhance response capabilities. Performance indicators such as evacuation times, traffic efficiency, and infrastructure reliability will guide data-driven recommendations for system improvements. The project’s outcomes will provide relevant stakeholders with actionable strategies to strengthen transportation resilience, safeguard mobility, and support sustainable infrastructure development.

Household adoption of green technologies, such as rooftop solar photovoltaic (PV) systems, heat pumps (HP), and electric vehicles (EV), plays a pivotal role in achieving net-zero greenhouse gas emissions by 2050. While existing studies have extensively examined factors influencing the adoption of individual technologies, limited research has focused on the co-adoption of multiple clean energy solutions.
This thesis aims to explore the combined influence of demographic, spatial, and housing factors on the adoption of clean energy technologies in Switzerland. The study will collect and analyze data at cantonal, municipal, and building levels, incorporating information on PV, HP, and EV adoption alongside relevant demographic and spatial factors. Regression-based models will be developed to assess the determinants driving the adoption of these technologies, offering insights into patterns of co-adoption and their underlying drivers. The research findings will contribute to a deeper understanding of household adoption behaviors and inform policy strategies for accelerating the clean energy transition in Switzerland.

If interested contact Palucci Matteo at: matteo.palucci@supsi.ch

The electrical grid is part of the critical infrastructure of every country. Its reliable and secure operation ensures resources for economy and society are always available. The energy transition, which is required among others, to reduce carbon emissions and contribute to more sustainable societies, requires integrating large amounts of distributed and renewable energy. While this constitutes a challenge by itself, it also makes operating the power grid increasingly difficult. Power now flows in “all” directions, in an ageing system, originally designed to provide energy from a few large producers hierarchically “top-down” to consumers.
To maintain reliable, cost-efficient operation of evolving grids, the AC optimal power flow problem (AC-OPF) must be solved more frequently. Traditional solvers, using Newton-Raphson methods, struggle with large-scale grids due to computational demands and convergence issues.
Neural power flow solvers address these challenges by shifting computational complexity from operations to training. Our GridFM–v0, leveraging graph neural networks (GNNs), is a foundation model designed as such a solver.
For GridFM–v0, grids are represented as graphs, where electrical buses are nodes with active power, reactive power, voltage magnitude and voltage angle as embeddings, and where transmission lines are the edges.
It is pre-trained using self-supervision, reconstructing masked node features of existing power flow solutions covering diverse grid topologies and load conditions.
A hybrid, physics informed loss function minimize reconstruction errors and ensures physical accuracy. This enables generalization across various scenarios and different grids including system security and optimal power flow tasks.
With the power grid being critical infrastructure, and the current geo-political situation, utilities are either very reluctant and/or legally bound to NOT make their grid data publicly available, which requires large amounts of training data to be produced synthetically. To address the data-challenge for power grid foundation models, the focus of the project is developing a unified tool for generating and accessing existing data from multiple sources. This tool will enhance the practical applicability of advanced models like GridFM–v0, enabling more realistic and actionable insights for modern power systems.

The electrical grid is part of the critical infrastructure of every country. Its reliable and secure operation ensures resources for economy and society are always available. The energy transition, which is required among others, to reduce carbon emissions and contribute to more sustainable societies, requires integrating large amounts of distributed and renewable energy. While this constitutes a challenge by itself, it also makes operating the power grid increasingly difficult. Power now flows in “all” directions, in an ageing system, originally designed to provide energy from a few large producers hierarchically “top-down” to consumers.
To maintain reliable, cost-efficient operation of evolving grids, the AC optimal power flow problem (AC-OPF) must be solved more frequently. Traditional solvers, using Newton-Raphson methods, struggle with large-scale grids due to computational demands and convergence issues.
Neural power flow solvers address these challenges by shifting computational complexity from operations to training. Our GridFM–v0, leveraging graph neural networks (GNNs), is a foundation model designed as such a solver.
For GridFM–v0, grids are represented as graphs, where electrical buses are nodes with active power, reactive power, voltage magnitude and voltage angle as embeddings, and where transmission lines are the edges.
It is pre-trained using self-supervision, reconstructing masked node features of existing power flow solutions covering diverse grid topologies and load conditions.
A hybrid, physics informed loss function minimize reconstruction errors and ensures physical accuracy. This enables generalization across various scenarios and different grids including system security and optimal power flow tasks.Building on this foundation, the focus of this project is to analyze the applicability of foundation models like GridFM–v0 for downstream tasks, including:

1. Optimal Power Flow (OPF): Evaluating GridFM–v0’s ability to improve computational speed and accuracy in solving optimal power flow problems under varying grid scenarios and conditions.
2. Cascading Failure Prediction: Exploring the use of GridFM–v0 to predict cascading failures by modeling how local disturbances propagate through the grid and identifying vulnerable areas.

The impact of climate change and extreme weather events is crucial to be assessed for the design and operation of climate-resilient energy systems, which involves scenario analyses. This project aims to leverage generative AI for the generation of climate scenarios and extreme events in an efficient manner, relieving the computational burden of solving partial differential equations. This project is a continuation of a previous semester thesis and the incoming student will start with an existing codebase. The student will be expected to improve the existing algorithms, add new ones to the current pool and perform sensitivity analyses. The student will gain exposure to climate data analysis, deep learning and generative AI in the context of energy systems analyses. A Masters thesis is preferred. Semester thesis can also be considered.

This master's thesis will tackle complex self-discharge and non-linear capacity fading effects for long duration storage technologies, especially batteries. This potentially impactful aspect is often underrated in linearized energy system optimization, operating on the large scale. Iterative approaches based on operational maps will be tested on different granularity of space and time aggregations on a regional energy model. The final goal is providing energy modelers with insights into storage fading and evaluating the impact on results and related decision-making process. Emerging technologies will also be included, since the activity is conducted in collaboration with the deep tech start-up Unbound Potential (UP), which is developing promising membrane-less flow batteries. The topic is particularly relevant to properly assess the performance of storage options, which will play an increasingly crucial role as sources of flexibility in the net-zero energy transition.

The rapid adoption of heat pumps (HPs), photovoltaics (PVs), and electric vehicles (EVs) is reshaping low-voltage (LV) distribution networks, posing challenges for Distribution System Operators (DSOs). Uncoordinated HP operations can worsen grid stress and drive costly reinforcements, while coordinated strategies can optimize grid performance and lower costs. This thesis addresses how DSOs can adapt to rising electricity demand by evaluating HP control schemes and pricing strategies to ensure sustainable and efficient grid management.

This master’s thesis aims to develop a comprehensive model to characterise the behaviours and potential of inverter-based resources (IBRs), including solar PV, wind, fuel cells, and battery storage, in supporting grid modernisation and enhancing energy security. By focusing on their role in providing ancillary services during grid disturbances, the project seeks to leverage IBRs to improve grid stability, energy efficiency, and resilience in a smart grid context. Key tasks include modelling IBR dynamics, integrating these models into broader power system frameworks, and analysing their capabilities to mitigate stability challenges in high renewable penetration scenarios.

In this master’s thesis, a new surrogate-assisted multi-level method for design under uncertainty is tested on a complex energy system. Indeed, sophisticated and large models are needed to capture the dynamic environment of power grids and understand how to make them resilient in the sustainable transition. The enormous complexity of the problem commonly forces oversimplification, just by accounting for a few scenarios among the many realizable socio-tecno-economic conditions. In this activity, an innovative, fast-scaling, yet accurate method able to approximate the variability of design variables and uncertain parameters interactions will be applied to the integrated Nexus-e model for Switzerland. The final goal is to disclose the most resilient investment and operative opportunities in the presence of a realistic variety of combinatorial uncertainties.

The energy transition is driving the deployment of distributed energy resources like photovoltaic panels, electric vehicles, and heat pumps. While beneficial, these technologies challenge power distribution grids (PDGs) that were not designed for such energy inputs. To address this, models of PDGs in Switzerland have been developed to assess the impact of various policies on the country's energy infrastructure. However, the complexity of the national system makes detailed modeling impractical. This thesis focuses on optimizing power system planning in representative PDGs and applying these findings to disaggregate results at a national level.

This internship, which can be extended with a master’s thesis at Swissgrid will tackle challenges at the heart of the energy transition: Current ancillary service markets need to be updated and expanded to accommodate distributed energy resources and address the increasing volatility in the power system due to renewable energy resources such as PV. The tasks will include implementing and testing market simulations, ad-hoc analyses and conceptual developments both within Swissgrid and in collaboration with other companies in the Swiss electricity sector. Ideally, the internship would start in February 2025 for 6-12 months. 

This master's thesis project aims to investigate the costs and benefits of local multi-energy systems in enhancing resilience and thereby providing a secure electricity and heat supply to end-users. To this aim, the in-house developed multi-energy system design optimization model needs to be enhanced such that component failures from multiple failure modes (such as random failures, from hazards such as snowstorms/windstorms) and variable repair time are characterized as uncertainty set and incorporated as a stochastic input to the design problem.