Strategies for real-time actuator decoupling in closed-loop MHD control operations

In many devices aiming at magnetic confinement of fusion relevant plasmas, feedback control of MHD instabilities by means of active coils is nowadays mandatory to ensure the robustness of high performance operational scenarios. Actuators involved in the control loop are often coupled in the sensor measurements and an optimal strategy for decoupling can be limited by the need of reducing as much as possible the cycle time of the control loop itself. RFX-mod is a medium size (R=2m, a=0.459m) device able of confining plasmas in both Tokamak and Reversed Field Pinch magnetic configurations. It is equipped with an advanced feedback system for field error and MHD control [1]. Actuators in this system are 192 active saddle coils entirely covering the plasma outer surface, while more than 600 magnetic sensors are included in the control loop, providing the operator with a challenging coupling situation. It is also important to stress the fact that the problem is intrinsically 3D, involving different non-axisymmetric contributions. All these characteristics can in principle be included in a (complex, frequency dependent) mutual coupling matrix that, when properly inverted and implemented in the control loop, should provide the necessary corrections to the actuator action. We will start by documenting the baseline situation in RFX-mod, where the Identity matrix is chosen to represent the simplest case of mutual coupling matrix. After that, implementation and use of decoupling matrices for the two limiting cases of zero and infinite frequency will be presented. The issue of properly modelling the problem will be also tackled by comparing results from white and black-box approaches.