Safety factor profile control with reduced central solenoid flux consumption during plasma current ramp-up phase using a reinforcement learning technique

Safety factor profile control via active feedback control of electron temperature profile during a plasma current ramp-up phase of a DEMO reactor is investigated to minimize the magnetic flux consumption of a central solenoid (CS) for wide range of q profiles. It is shown that q profiles with positive, weak and reversed magnetic shear can be obtained with the resistive flux consumption less than 60% of the empirical estimation which is calculated using the Ejima constant of 0.45. For the optimization of the target electron temperature profile and feedback gain to control electron heating power, reinforcement learning technique is introduced. One important feature of the system trained by reinforcement learning is that it can optimize the target electron temperature adaptive to the present status of a plasma. This adaptive feature of the reinforcement learning enables to control q profiles even in the case that an effective charge profile is randomly modified and it is not measured directly.