Nuclear fusion is a promising long-term solution for large-scale and low-carbon energy production, relying on the confinement of high-temperature plasmas in magnetic devices such as tokamaks. Achieving and sustaining high fusion performance requires operation close to physical and technological limits, where plasma behavior becomes highly nonlinear and sensitive to perturbations. In this context, accurate real-time observation and control of plasma kinetic profiles, such as the electron density, temperature, and current density, are essential to maximize performance, ensure operational safety, and avoid instabilities and limit violations.
Among kinetic quantities, the electron density plays a central role in determining plasma confinement, heating efficiency, exhaust conditions, and proximity to the so-called density limit. Density levels and profiles directly affect wave absorption for auxiliary heating and current drive systems, impurity transport, and access to high-performance regimes. However, real-time measurements are inherently limited by diagnostic availability, signal-to-noise ratio, and differing spatial and temporal resolutions, motivating the use of observers to reconstruct the plasma state in real time.
This thesis focuses on the real-time observation and control of the electron density profile on the TCV tokamak using a newly upgraded RAPDENS observer, an adaptive control-oriented model of particle transport coupled with an Extended Kalman Filter formulation. A novel multi-rate estimation framework is employed to fuse high-frequency interferometry measurements with lower-rate Thomson scattering data, as well as integrated with local in (R,Z) space measurements, enabling improved spatial and temporal reconstruction of the density profile while ensuring robustness to diagnostic offsets and data unavailability or data outliers.
This new observer framework is tested and validated along four main topics: control of the line-integrated density within the last closed flux surface (LCFS), differentiating with the line-integrated density in the scrape-off layer (SOL), control of a local density at a given flux surface within the LCFS, control of the edge density at high performance and near the density limit, and finally density control while the diagnostic used in standard real-time control was absent.
First, the RAPDENS observer is leveraged to control the plasma density (inside the LCFS) in detachment studies with divertor legs of varying length. The observer enables feedback control that is decoupled from scrape-off layer contributions, even in cases with dominant integrated density from the SOL along a far-infrared interferometer line of sight. This allows detailed physics studies and reproducible density trajectories across different divertor configurations, impurity-seeded conditions, and heating regimes.
The observer is then employed for local density profile control in Electron Cyclotron Resonance Heated (ECRH) and Neutral Bea
