Research Projects

My research focuses on the interface of stochastic modelling, optimisation, and control, applied to large-scale systems such as energy infrastructures and financial networks. A central theme is understanding how uncertainty propagates through complex, interconnected structures, and how rigorous mathematical design can mitigate risk while improving stability, interpretability, and efficiency.

The projects presented here reflect this perspective. They range from methodological advances in stochastic and distributed control to domain-driven applications in smart energy systems, credit risk, and emerging financial technologies. Together, they outline a trajectory that connects theoretical innovation with practical impact across critical infrastructures.

NoteSelected Awards & Visits
  • 🏆 Paul M. Frank Award — IFAC SAFEPROCESS 2018
  • 🌏 Research visits — AIP Tokyo (2018, 2019); CORE Louvain-la-Neuve (2016); ETH Zürich IfA (2012–2013)
  • 🎓 Scholarships — AIP Tokyo (2018, 2019); Politecnico di Milano MSc; ETH Zürich MSc research visit

Current Research Focus

Scenario Convexification of Stochastic Nonconvex Programs

Description

This project addresses one of the central difficulties in stochastic optimisation: nonconvex programmes with chance or expectation constraints. Such problems arise naturally in energy, finance, and engineering applications, yet their scenario approximations often suffer from looseness and unpredictable duality gaps.

We reformulate the scenario problem using a set-valued perspective, where feasibility is represented as the Minkowski sum of per-scenario sets. By combining Aumann expectations with a law of large numbers for random sets, we establish convergence properties of these sums. Applying a Shapley–Folkman convexification argument, we derive explicit bounds on the duality gap.

For chance-constrained problems, the duality gap is shown to be uniformly bounded by a constant depending only on dimension. For expectation-constrained problems, the gap vanishes at rate O(1/N) with the number of scenarios. Numerical experiments illustrate these asymptotic behaviours and reveal the distinct roles of chance versus expectation constraints in shaping feasible regions.

Research significance

These results offer one of the first rigorous analyses of scenario methods in nonconvex optimisation. They demonstrate that despite nonconvexity, scenario approximations can achieve provable stability, either through bounded or vanishing duality gaps. This advances the mathematical foundations of risk-aware optimisation with implications for portfolio design, energy systems, and engineering control.

  • Theme: Stochastic optimisation, scenario methods, convexification
  • Institution: Joint work with Akiko Takeda (University of Tokyo) and Takafumi Kanamori (Institute of Science Tokyo)
  • Status: In preparation for submission to SIAM Journal on Optimization (SIOPT)
  • Keywords: Duality gap, random sets, Shapley–Folkman theorem, chance constraints, expectation constraints

Outcomes

  • Introduced a scenario convexification framework for analysing stochastic nonconvex programmes.
  • Derived dimension-dependent uniform bounds for chance-constrained problems.
  • Proved vanishing duality gaps for expectation-constrained problems with rate O(1/N).
  • Provided numerical evidence supporting theoretical asymptotics.

Highlighted outputs

  • Working paper: Vanishing Duality Gap in Uncertain Nonconvex Problems (with A. Takeda and T. Kanamori).
  • Target journal: SIAM Journal on Optimization (SIOPT).

Anchored Kalman Filtering for Lifetime PD

Description

Lifetime probability of default (PD) estimation is central to credit impairment modelling under IFRS 9 and CECL. A recurring challenge is that macroeconomic forecast errors accumulate when projecting transition matrices over long horizons, leading to volatile and procyclical PD term structures.

This project reformulates the problem in a state–space framework, where term structure dynamics are estimated with a Kalman filter. We show that the standard filter leaves non-vanishing variability, failing to stabilise long-run PD paths. To address this, we introduce an anchored observation model that incorporates a neutral long-run economic state into the filtering process. This adjustment ensures asymptotic stochastic stability, providing convergence in probability of the lifetime PD term structure.

Research significance

This project advances the reliability of credit risk modelling under IFRS 9 and CECL by addressing one of its core weaknesses: unstable and procyclical lifetime PD term structures. By embedding PD estimation in a state–space framework and introducing anchored observation models, the work provides a mathematically rigorous solution that ensures stochastic stability. Beyond theory, the approach directly supports banks and regulators in producing smoother, more interpretable projections of credit risk, reducing model risk and enhancing comparability across economic cycles.

  • Theme: Credit risk, IFRS 9 / CECL, model stability
  • Institution: Independent research (in collaboration with industry applications)
  • Status: Working paper, simulation package, GitHub repository
  • Keywords: Risk modelling, stochastic stability, state–space methods

Code arXiv

Outcomes

  • Demonstrated that forecast noise can be substantially reduced by anchoring the observation model.
  • Derived conditions for stochastic stability of PD term structures under filtering.
  • Developed a simulation package illustrating effects on synthetic corporate portfolios.
  • Produced smoother, more interpretable PD projections, with direct implications for regulatory model risk management.

Highlighted outputs

  • Working paper: Anchored Kalman Filtering for Lifetime PD Models under Forecast Uncertainty (submitted to the Journal of Credit Risk).
  • GitHub repository with simulation package.
  • arXiv preprint: arXiv:2509.10586.

Past Research Projects

ERSAS — Energy System Integration — Postdoc Project

Description

The ERSAS project (Efficient, Reliable, Sustainable and Socially Acceptable Energy System Integration) developed new approaches to the coordinated operation of electricity, gas, and heat networks. The central idea was that multi-carrier infrastructures — such as Combined Heat and Power (CHP) units, hybrid heat pumps, Power-to-Gas, and Gas-to-Power conversion — can significantly improve efficiency and flexibility if supported by the right control strategies and market mechanisms.

My work focused on mathematical modelling, stochastic optimisation, and distributed control design. I studied how uncertainty in renewable generation and end-user demand propagates across coupled infrastructures, and how incentive schemes and control algorithms can align decentralised actors with system-level goals. This included developing model predictive control (MPC) formulations that balance flexibility, efficiency, and reliability, while explicitly accounting for financial and operational risks.

  • Theme: Electricity–gas–heat coupling, incentives, and control
  • Institution: University of Groningen (with partners: Alliander, Enexis, TNO)
  • Funding: Netherlands Organisation for Scientific Research (NWO)
  • Period: Sep 2018 – Dec 2019

Outcomes:

  • Demonstrated how coupling gas and electricity systems can improve integration of renewable energy while maintaining security of supply.
  • Proposed stochastic MPC frameworks with risk measures (e.g., CVaR) to better manage demand flexibility under uncertainty.
  • Designed distributed algorithms for building-to-grid integration and frequency support services.
  • Provided simulation tools to evaluate multi-carrier energy scenarios, supporting both academic and industrial partners.

Highlighted Publications:

  • J. Venkatasubramanian, V. Rostampour, T.S. Badings, T. Keviczky. Stochastic MPC for Energy Management in Smart Grids with CVaR as Penalty Function. IEEE PES ISGT-Europe, 309–313, 2020. DOI: 10.1109/ISGT-Europe47291.2020.9248769
  • V. Rostampour, T. Keviczky. Demand Flexibility Management for Buildings-to-Grid Integration with Uncertain Generation. Energies, 13(24):6532, 2020. DOI: 10.3390/en13246532
  • V. Rostampour, T.S. Badings, J.M.A. Scherpen. Buildings-to-Grid Integration with High Wind Power Penetration. CDC, 2976–2981, 2019. DOI: 10.1109/CDC40024.2019.9030242
  • T.S. Badings, V. Rostampour, J.M.A. Scherpen. Distributed Building Energy Storage Units for Frequency Control Service in Power Systems. IFAC SGRES, 52(4):228–233, 2019. DOI: 10.1016/j.ifacol.2019.08.190

ATES-SG — Smart Thermal Grids with Aquifer Storage — PhD Project

Description

ATES-SG studied how Aquifer Thermal Energy Storage (ATES) can be scaled in dense urban districts while maintaining efficiency and minimising interference between neighbouring systems. We modelled interactions among hot/cold wells, uncertain heat-cooling demand, and conversion assets to design distributed MPC and planning tools that coordinate operation at district scale. The work connected subsurface physics with network-level control and market incentives, showing how seasonal storage can reduce emissions and peak loads when orchestrated with building and grid objectives.

  • Theme: Urban energy and seasonal storage
  • Institution: Delft University of Technology (with Waternet, Provincie Noord-Holland, Tauw BV, STW/TTW, Priva, DWA)
  • Funding: Netherlands Organisation for Scientific Research (NWO)
  • Period: Sep 2014 – Aug 2018

Outcomes

  • Frameworks for distributed energy management of multiple ATES sites under uncertainty.
  • Evidence that coordination mitigates thermal interference, improving seasonal performance factors.
  • MPC formulations linking building comfort, storage dynamics, and market signals.
  • Open simulation setups that supported partner evaluations and scenario studies.

Highlighted publications

  • V. Rostampour, T. Keviczky. Probabilistic Energy Management for Building Climate Comfort in Smart Thermal Grids with Seasonal Storage Systems. IEEE Trans. Smart Grid, 10:3687–3697, 2018. DOI: 10.1109/TSG.2018.2834150 arXiv:1611.03206 PDF
  • V. Rostampour, O. t. Haar, T. Keviczky. Distributed Stochastic Reserve Scheduling in AC Power Systems With Uncertain Generation. IEEE Trans. Power Systems, 34:1005–1020, 2018. DOI: 10.1109/TPWRS.2018.2878888
  • V. Rostampour, T. Keviczky. Demand Flexibility Management for Buildings-to-Grid Integration with Uncertain Generation. Energies, 13(24):6532, 2020. DOI: 10.3390/en13246532
  • V. Rostampour, R. Ferrari, A. Teixeira, T. Keviczky. Differentially-Private Distributed Fault Diagnosis for Large-Scale Nonlinear Uncertain Systems. IFAC SAFEPROCESS, 51(24):975–982, 2018. DOI: 10.1016/j.ifacol.2018.09.703
  • V. Rostampour, D. Adzkiya, S. Soudjani, B. De Schutter, T. Keviczky. Chance-Constrained MPC for Stochastic Max-Plus Linear Systems. IEEE SMC, 3581–3588, 2016. DOI: 10.1109/SMC.2016.7844789
  • V. Rostampour, T. Keviczky. Energy Management for Building Climate Comfort in Uncertain Smart Thermal Grids with ATES. IFAC World Congress, 50(1):13698–13705, 2017. DOI: 10.1016/j.ifacol.2017.08.2170
  • V. Rostampour, M. Bloemendal, T. Keviczky. MPC of GSHP coupled with ATES System in Heating and Cooling Networks of a Building. IEA Heat Pump Conference, Rotterdam, 2017.

MOVES — MSc Thesis — ETH Zürich & Polimi

Description

As part of the MOVES project (Modeling, Optimization, and control for the integration of Electric vehicles and renewable Sources), my Master’s thesis focused on tractable formulations for reserve scheduling in power systems with uncertain renewable generation. This early work explored how stochastic optimisation and probabilistic certificates could be applied to system-level scheduling problems, anticipating challenges in high-renewable grids. The project provided me with first-hand exposure to uncertainty modelling in energy systems and laid the foundation for my later research in stochastic MPC and distributed optimisation.

  • Theme: Reserve scheduling with uncertain generation
  • Institution: Automatic Control Laboratory (IfA), ETH Zürich (in collaboration with Politecnico di Milano)
  • Advisors: Dr. Kostas Margellos, Dr. Maria Vrakopoulou, Prof. John Lygeros, Prof. Maria Prandini
  • Period: MSc Thesis, completed April 2013

Outcomes

  • Developed reserve scheduling formulations that explicitly incorporate uncertainty in renewable generation.
  • Demonstrated tractable probabilistic guarantees for system security under renewable variability.
  • Early contribution bridging optimisation theory and practical power system operation.

Highlighted publications

  • V. Rostampour, K. Margellos, M. Vrakopoulou, M. Prandini, G. Andersson, J. Lygeros. Reserve Requirements in AC Power Systems With Uncertain Generation. IEEE ISGT-Europe, 1–5, Copenhagen, 2013. DOI: 10.1109/ISGTEurope.2013.6695354
  • K. Margellos, V. Rostampour, M. Vrakopoulou, M. Prandini, G. Andersson, J. Lygeros. Stochastic Unit Commitment and Reserve Scheduling: A Tractable Formulation with Probabilistic Certificates. ECC, 2513–2518, Zürich, 2013. DOI: 10.23919/ECC.2013.6669532
  • MSc Thesis: Tractable Reserve Scheduling Formulations for Power Systems with Uncertain Generation (2013). PDF

Applied Research & Entrepreneurial Initiatives

ZWMCo — Tokenised Hedge Fund with DeFi Risk Framework

Description

ZWMCo explores how quantitative risk modelling can be embedded into decentralised finance. The project envisions a tokenised hedge fund where investors contribute to a managed pool and receive a native token representing their share. Unlike many existing DeFi products, ZWMCo emphasises transparent risk management, leveraging methods from quantitative finance such as liquidity stress testing, stochastic control, and scenario analysis.

The system integrates two main components:

  • Algorithmic trading (AVS-based): A suite of strategies built on Adaptive Volatility Scaling (AVS) principles. These strategies dynamically adjust exposure across crypto assets, balancing risk and return under changing volatility regimes.
  • LSDx framework: A research-driven risk and pricing model for liquid staking derivatives (LSDs). The framework establishes valuation signals, stability bounds, and margin rules to reduce systemic fragility in LSD markets, providing an institutional-grade lens for DeFi liquidity products.

Research significance

ZWMCo illustrates how academic expertise in stochastic modelling, optimisation, and risk pooling can extend into Web3 ecosystems. By combining AVS-based algorithmic trading with an LSDx valuation framework, the project bridges quantitative finance with decentralised tokenomics, offering a path to sustainable, risk-aware DeFi investing.

Theme: Decentralised finance, algorithmic trading, liquidity risk
Institution: Independent initiative (Zillions Wealth Management Co.)
Status: Concept development, white paper drafting, prototype trading bot
Keywords: DeFi, staking derivatives, risk management, tokenisation

Outcomes

  • Drafted a technical white paper on LSDx risk and pricing frameworks.
  • Designed a prototype trading bot implementing AVS-based strategies with monitoring features.
  • Developed a tokenomics model with presale and buy-back mechanics.
  • Planned GitBook-based documentation for transparency and community adoption.

Highlighted outputs

  • Technical white paper on LSDx (in preparation).
  • AVS trading bot prototype (private repository).
  • Draft GitBook documentation for public launch.

ChargePool — Behavioural Energy Pricing for EV Charging

Description

ChargePool tackles one of the major challenges for electric vehicle adoption: the uncertainty and volatility of charging costs. The project builds on my expertise in stochastic optimisation and risk pooling to design a system where EV users can “lock in” charging prices over a future horizon. By aggregating user demand and leveraging hedging strategies, the platform provides predictable energy costs while balancing risk across the pool.

The innovation lies in combining behavioural modelling with financial risk techniques: users interact through a simple app, but underneath, advanced optimisation allocates risk, manages variability, and aligns incentives. Beyond pricing, ChargePool also integrates route planning, station availability, and points of interest, turning EV charging into a more seamless mobility experience.

Research significance

ChargePool represents a direct application of stochastic modelling and incentive design to consumer-facing energy systems. By bridging academic methods and entrepreneurial deployment, it illustrates how optimisation and risk management can create new markets for flexibility and predictability in the energy transition.

Theme: Energy systems, behavioural economics, pricing optimisation
Institution: Independent initiative (with prospective industry partners in Switzerland)
Status: Proof-of-concept (submitted for BRIDGE funding), prototype app in development
Keywords: EV charging, risk pooling, smart pricing, behavioural modelling

Outcomes

  • Designed a risk pooling engine that stabilises charging costs for users.
  • Developed a prototype app for route planning and charging station selection.
  • Initiated discussions with Swiss energy and mobility partners (e.g., GoFast, Swisscard, Energy360).

Highlighted outputs

  • Early prototype code and simulation results (private GitHub repository).