Nonlinear dynamic modeling for control of fusion devices Alessandro Beghi, Mario Cavinato and Angelo Cenedese* Abstract— In this paper the issue of modeling for control design with application to fusion devices is discussed. Given the difficulty to provide analytical solutions to the equations describing the system dynamics, the use of numerical tools for evaluating different control architectures is required. Although standard tools for Computer Aided Control System Design (CACSD) can be usefully employed for both analyzing the fu- sion device dynamics and designing control laws, dedicated tools are required to adequately describe the complex interactions among the different system components. In this paper the use of the nonlinear equilibrium code MAXFEA to support the control design task is described. MAXFEA is a finite element code able to produce quite good approximations of the plasma boundary location and shape, together with internal distributions of current and magnetic fields, and other plasma features. The code provides the simulation of the plasma dynamics, while all the other elements (diagnostics, controller, actuators) in the control loop system can be modeled independently and integrated in the code as external modules, thus making it a candidate tool for Software-in-the-Loop solutions. I. INTRODUCTION One of the most promising and viable approaches to produce energy from nuclear fusion reactions, on Earth and in a controlled way, is to resort to the magnetically confined plasmas, which yield a relevant rate of nuclear reactions while kept inside a metallic toroidal chamber at extremely low pressures and high temperatures. The equation governing the equilibrium of a magnetically confined plasma is the Grad-Shafranov equation, a two dimensional nonlinear elliptic partial differential equation [1] that basically states the equilibrium of a plasma column with an external magnetic field, in hypothesis of axisymmetry. The toroidal plasma current density profile J p (ψ) is given as a function of the poloidal flux ψ by the following J p (ψ)= L(ψ)= -µ 0 r dp(ψ) dψ - 1 r f (ψ) df dψ , (1) with L being the operator L = ∂z 1 r ∂z + ∂r 1 r ∂r  . (2) The profile is thus parameterized according to the two functions p(ψ) and f (ψ), respectively the pressure profile and the poloidal current density flux function. *Corresponding author. A. Beghi is with Department of Information Engineering, University of Padova, Padova, Italy beghi@dei.unipd.it M. Cavinato is with Consorzio RFX - EURATOM Fusion Association, Padova, Italy mario.cavinato@igi.cnr.it A. Cenedese is with the Department of Engineering and Management, University of Padova, Vicenza, Italy angelo.cenedese@unipd.it Fig. 1: Control Loop scheme. The open loop is the cascade of power supply, plasma-vessel system, and diagnostics. Solving the Grad-Shafranov equation means to compute the magnetic flux distribution for a given external coil current configuration, whence the plasma boundary (its flux value, location, and shape) is determined. To compute a boundary for a plasma in equilibrium is known as the free boundary problem. Actually, the nonlinear nature of the problem arises from the parametrization chosen for p(ψ) and f (ψ) and the free-boundary condition itself. Unfortunately, in general this does not offer an analytical solution so that a common way to solve the problem is to employ numerical procedures, as finite element or finite difference methods, implemented in equilibrium codes. At the same time, the magnetic control of a fusion device is required for the control of the macroscopic characteristics of the plasma mass (shape, position, modes, current) both to drive the plasma during the various phases of the discharge and to counteract disturbances and internal instabilities, and is exerted by acting dynamically on the magnetic field produced by external coils surrounding the vacuum chamber. The use of equilibrium codes is of fundamental importance not only for the inherent level of detail, but also for the perception of the dynamics of the complete system obtained through the nonlinear simulation. In fact, the complete closed loop system can be simulated in the equilibrium code: The vessel-plasma system, the diagnostics (output measure- ments), the actuators (input signals). The power supply block and the diagnostics can be easily understood and their accurate modeling poses stringent constraints on system controllability [2]. The vessel-plasma system basically refers to the complex interaction among the effects of the active coil currents, the eddy currents flowing in the machine structure (passive metallic structures, plasma facing components, ...), and the plasma, to produce the output quantities measured by the diagnostic sensors. In particular, the possibility of performing long time steady state simulations, or even simulations of different phases of the plasma discharge with the relevant transitions, represents Proceedings of the 47th IEEE Conference on Decision and Control Cancun, Mexico, Dec. 9-11, 2008 WeB18.3 978-1-4244-3124-3/08/$25.00 ©2008 IEEE 3133