Period: 01/04/2024 – 31/07/2024

SECTION 1 – GENERAL TRENDS OF THE PROJECT

Brief summary of the project

With reference to the work packages/tasks reported in Table 1, the following progress has been made during the considered timeframe:

• WP0 – Task 0.1. both the parameter varying linear model and the Simulink version of the plasma equilibrium code CREATE-NL have been made available to the control design tasks. They will be finalized and delivered in their final version in the next bimester;
• WP1 – Task 1.3. First proposal of a data-driven magnetic control system based on a Deep Deterministic Policy Gradient (DDPG) technique. In following bimesters, the parameter varying linear model developed by Task 0.1 will be used for the training of this control system.
• WP2 – Task 2.1. Two neural networks-based solutions, one based on Multilayer Perceptrons (MLP) and another based on Extreme Learning Machines (ELM) have been  developed to estimate the movement of the plasma along the unstable mode.
• WP2 – Task 2.2. This task has been anticipated, since it was originally assumed to start in the next bimester. A first proposal of a fully data-driven Vertical Stabilization (VS) system that relies on the neural network estimator currently developed by Task 2.1 has been made.

Table 1 - TRAINER work packages and tasks

Names of the operational units involved in the implementation of the project

• Research Unit (RU) University of Naples “Federico II” – led by the Principal Investigator, Prof. Gianmaria De Tommasi
• RU University of Naples “Parthenope” – led by Prof. Marco Ariola

The project activities of the considered timeframe were mainly related to:

• development and use of the plasma models for controller design (i.e., agent training) and validation;
• development of data-drive magnetic control system based on three DDPG agents dedicated to the basic magnetic control problems
• Parthenope RU is purchasing the following workstation to perform the training of the data-driven controllers
  • Dell Alienware AURORA R16 with Intel Core i9-13900KF (24 cores, 32 threads), NVIDIA GeForce RTX 4090 (16384 CUDA cores, 24GB RAM) and 64GB of RAM.
• publication of the following papers/presentations at International conferences:
  • [P1] De Tommasi, L. E. di Grazia, S. Dubbioso, F. Fiorenza, D. Frattolillo, S. Inoue, M. Mattei, A. Pironti, H. Urano, “Control of elongated plasmas in superconductive tokamaks in the absence of in-vessel coils,” Nuclear Fusion, vol. 64, no. 7,
    pp. 076005, July 2024 (about the reference functional architecture for plasma magnetic control and model validation on JT-60SA equilibria).
  • [P2] E. di Grazia, C. Vincent, M. Mattei, F. Felici, L. Kogan, A. Mele, “Iterative Learning Optimisation and Control of MAST-U Breakdown and Early Ramp-Up Scenarios,” 10^th International Conference on Control, Decision and Information Technologies (CoDIT 2024), Valletta, Malta, July, 2024 (about validation of plasma modeling tools on the MAST-U spherical tokamak).

Description of the carried out activities which are in compliance with the DNSH, Open Access principles as well as with gender, generational principles and with those of Equal opportunities

In the considered timeframe, the journal paper [P1] has been published as Open Access. As for DNSH and gender, generational principles, and with those of Equal opportunities no specific compliance have been experienced, in this timeframe.

Description of the actions aimed at informing and disseminating knowledge

The main activities performed in the considered timeframe are:

1. participation to CoDIT 2024 conference to present the contribution [P2];
2. update of the contents of the website https://trainer.dieti.unina.it/.

SECTION 2 – PROGRESS OF ACTIVITIES

Detailed description of activities carried out by each operational unit with a focus on the timeframe for their implementation

In the considered timeframe, the Federico II RU has carried out two main activities:

1. finalization of the plasma modelling tools (Task 0.1);
2. development of a first fully data-driven architecture for an Extremum-Seeking based (ES-based) plasma Vertical Stabilization (VS) system.

As far as item i) both the Simulink based nonlinear equilibrium code and the parameter varying linear systems have been finalized and tested on different machines, such as JT-60SA and ITER. The validation carried out for the ITER Plasma Control System (PCS) will be reported in a contribution currently under preparation planned to be presented at the next Symposium on Fusion Technology (SOFT) Conference, that will be held in Dublin in September 2024 (https://soft2024.eu/conference/). The delivery of modelling tools is envisaged at the beginning of the next bimester.

Concerning item ii), the control architecture reported in Figure 1 has been tested and validated for the ITER case study, by using the modelling tools include in item i). The proposed architecture exploits a neural network estimator of the plasma movement along the unstable mode. Such estimation is be fed back into a stabilizing ES algorithm [R1].  

Extensive simulations have been made, and now the Federico II RU started a collaboration with the Swiss Plasma Center (SPC) of the École Polytechnique Fédérale de Lausanne (EPFL), to test such an approach on the TCV tokamak.

Figure 1 - Block diagram of the proposed ES-based VS system. It includes the event-driven control gain adaptive mechanism and a Neural Network used to reconstruct the plasma movement along the unstable mode.

In the same timeframe, Parthenope RU has worked on the development of a data-driven control system based on DDPG agents. The control architecture reported in Figure 2 has been tested and validated for a DEMO case study with a plasma limited, by using the CREATE-NL modelling tools. The proposed architecture is based on three different DDPG agents performing the control tasks of the magnetic confinement. The coupling phenomena among the tasks and the controllers are taken into account by developing a proper training procedure. This modular approach is exploited to simplify the training procedure and to reduce the computational cost needed to perform the training of the overall control system.

Figure 2 - Block diagram of the proposed DDPG-based magnetic control system.

References

[R1] A. Scheinker, M. Krstić, “Model-free stabilization by extremum seeking”, Springer, 2017.

Description of potential changes to what has been originally approved mentioning the impacts on the aim of the intervention, on the achievement of intermediate and long- term goals, on the proposed actions for improvement

Activity on Task 2.2 have been anticipated. Such a minor change in the original schedule will not affect the overall project objectives.

Description of potential challenges encountered and of the proposed actions for improvement

With regard to the specific timeframe, no major challenges or issues have been encountered.

Brief description of potential publications

The following paper that describes a fully data-driven ES-based VS is in preparation and it is planned to be submitted to Expert Systems With Applications (https://www.sciencedirect.com/journal/expert-systems-with-applications):

• Dubbioso, A. Jalalvand, J. Wai, G. De Tommasi, E. Kolemen, “Model-free Stabilization via Extremum Seeking using a cost neural estimator.”

while several contributions to the SOFT 2024 Conference (September 2024) are under preparation, concerning the further validation of the CREATE modelling tools on ITER and Italian Divertor Test Tokamak (DTT), as well as a possible deployment of a simplified version of the ES-based VS at the TCV tokamak in Lausanne, that will allow to experimentally validate the approach:

• Dubbioso, G. De Tommasi, C. Galperti, F. Felici, S. Marchioni, A. Mele, A. Merle, A. Tenaglia, “Simulation validation of an Extremum Seeking-based Vertical Stabilization system for TCV.”
• Dubbioso, D. Ottaviano, F. Fiorenza, N. Ferron, G. Manduchi, R. Ambrosino, G. De Tommasi, “Rapid prototyping of control modules for the DTT Plasma Control System.”
• Mattei, R. Ambrosino, M. Ariola, G. De Tommasi, P. de Vries, A. Pironti, L. Zabeo, “Recent Developments in ITER Magnetic Control Algorithms.”

SECTION 3 – COMMON INDICATORS

Below the updates on the indicator RRFCI 8 – “Number of researchers who work in research centres which are recipients of financial support (women; men; non-binary)” – as per the description in the guidelines included in the n. 34 MEF notification from the 17^th of October 2022

SECTION 4 – PREDICTIVE ANALYSIS AND FINAL COMMENTS

Below it is provided a description of the forecast scenario on the development of the project, any potential change which is deemed necessary for the future as well as comments on the document.

Predictive analysis

The envisaged plasma modelling tools have been almost finalized, and they will be delivered in the next bimester. The data-driven control development has started as planned: in particular Parthenope RU started to investigate on the solution of the basic magnetic control problems by means of DDPG agents, while a first fully data-driven architecture for an ES-based VS system has been proposed by Federico II RU. In the next bimesters it is planned to test such a
data-driven VS  on the TCV tokamak.

Final comments

The project is progressing as planned. The task related to the development of the ES-based VS has been slightly anticipated, but this will not affect the overall outcome of the project.

ATTACHMENTS

The below documents are also attached to the technical – scientific report:

Att.1 – Declaration of compliance with DNSH principle and compliance with other principles as per the Environment code