Assessment of reflectometry diagnostics for DEMO

Abstract:
DEMO will be the first prototype of a fusion power plant. Unlike experimental tokamaks, only the necessary diagnostics for machine protection and plasma control will be implemented. One of the fundamental measurements is the position and shape of the last closed magnetic surface, typically measured with the magnetic diagnostics. One of the major issues of its implementation in DEMO is the large integration drifts that can occur during the operation due to the high levels of radiation. This can lead to a wrong plasma position estimation, putting the operation at risk.

The prime candidate to complement or substitute the magnetic diagnostics in DEMO is the microwave reflectometry. By sweeping the frequency of the probing beam, microwave reflectometry is capable of measuring the electron density profile. As the density is directly linked to the magnetic flux surfaces these measurements give access to the magnetic configuration, providing its local radial position. The O-mode propagation is independent from the magnetic field, being ideal for replacing the magnetic diagnostics.

The DEMO plasma position reflectometer (DEMO PPR) consists a system of multi­ reflectometers distributed poloidaly along the wall at different positions that will provide the separatrix reconstruction. The optimization of such system requires the simulation of the measurement process for different poloidal views, emitting angles, antenna assemblies and plasma configurations. The final system must be optimized for the operation scenario and be stable under the possible deviations to its equilibrium that can occur during the discharges. This includes the plasma displacement, turbulence or MHD activity. For now, the DEMO PPR is in an early development stage and there are many questions that need to be investigated before reaching its final design.

In this work we study the process of optimization of PPR systems with a general approach, taking into account the future changes in the geometry and plasma scenario. The important variables of a general multiple reflectometers system were identified and the techniques and the procedure to optimize it were developed. The simulation of such systems is in general a complex task that requires the definition of several different regions of interest and testing different antenna models and plasmas, which is a very demanding task from the computational point of view and of necessary time to build the simulation scripts. For this reason, we develop the structure of a high-level framework for multiple reflectometry simulations that is capable of automatizing all the simulation process of a multiple reflectometers system for the REFMUL* codes, a family of full­ wave FDTD codes that has been used for reflectometry simulations. The user defines the configuration files of the system geometry and plasma, the probing bands and the dependence between the main variables of the problem. A script creates all the necessary models and scripts that to run all the simulations in the HPCs.

Using the developed framework, we optimized the DEMO PPR system using the official DEMO scenario from EUROFUSION. We started by defining 100 different positions around the tokamak and testing two different configurations. In the first one, the antennas were aligned perpendicularly with the wall. This configuration has advantages from the point of view of the implementation of the antennas. However, the results shown that there are positions in the top of the machine and in the divertor region that have a very poor measurement performance and in some cases the signal is totally lost. In the second configuration, the antennas were aligned perpendicularly to the separatrix. In this case, since the direction of the probing beam is approximately parallel to the density gradient, a better measurement performance is expected. The results confirmed this principle, improving the results in several positions. At the divertor region, some of the positions continued to have a poor measurement performance, being necessary to sweep the probing angle to verify if there is an optimized configuration.

One of the problems associated with the optimization is that is necessary to extract the round trip group delay and calculate the amplitude of the detected signal for many different configurations. The analysis of the simulation results requires the manual adjustment of the data analysis parameters, as the filter cutoff frequency or the signal delay. Using the principle that a slow varying group delay has a minimum standard deviation if it is well filtered, we developed an automatized version of the the 1/Q detection, designed as IQA method. With this technique, it was possible sweeping the probing angle at all the positions in an acceptable time and select the optimized configuration. The maximum average detected amplitude shown to be a good selection criteria for the optimized configuration. The results show that, with the exception of some positions in the divertor region, there is an optimized configuration with low position error {<1 cm) and the power losses minimized.

With the optimized configuration, the stability of the system was tested for plasma displacements of 5 (reference case) and 15 cm (limit case). The results show that, in contrast to the results of the configuration with the antennas perpendicular to the separatrix, the system is stable for 5 cm plasma displacements in different directions (0, 90, 180 and 270 degrees). For displacements of 15 cm, the positions at the top of the machine can reach errors in the order of the minimum error requirement (1 cm).

The effect of turbulence in reflectometry measurements was studied in one gap of the equatorial region, in the high field side. Due to the lack of information on the turbulence properties of DEMO plasmas, the fluctuations were defined with an analytical model. A Kolmogorov-like spectrum was used to generate 400 random plasma samples for 16 different levels of amplitude (1-16%), compatible with the order of values observed in the experiments. Using the IQA algorithm, the principal statistical parameters were calculated. The results show that for higher levels of turbulence ($>5%$), the mean position error becomes negative due to the change of the effective cutoff position. This effect occurs for all the frequencies, leading to an accumulative error that can affect the position measurement in the order of accuracy requirements. In order to prove the reliability of the entire system, it is necessary to apply the same procedure to the other positions of the system, which requires a huge amount of computation time on HPCs to be done.

The techniques and the algorithms developed in this work can be applied in other processes which involve the analysis of a high number of reflectometry simulations, including studies with other reflectometry techniques.