Advanced vertical control for tokamak reactors with double-null divertors

双零导流器是紧凑型托卡马克反应堆的一种极具吸引力的排气解决方案，但其成功与否取决于实现垂直等离子体位置控制，该控制既鲁棒又精确，并且使用一些非磁性状态估计。一个自然的选择是感应两个分流器之间的热流不平衡，因为这个数量正是我们希望最小化以确保平等的功率共享。等离子体垂直运动和被检测热流之间的时间延迟是该控制概念中的一个重要参数，其中一部分是由刮擦层（SOL）中的瞬变造成的。如果信息沿SOL快速平流，那么热通量就成为等离子体位移的一个很好的估计，但如果它很慢（与垂直振荡时间标度相比），那么闭环控制就变得更具挑战性。目前尚不清楚反应堆的这种特征SOL时间是多少，但通过设计一种在这种情况下所需的先进垂直控制方案，似乎可以谨慎地减少它“缓慢”的可能性。我们建议创建一个模型预测控制算法，该算法利用基于时间依赖的SOL行为的快速简化/代理模型，根据最近的测量，实时预测系统的未来演变，并相应地在垂直控制线圈上启动。该模型将通过mat - u数据和仿真进行验证，并通过计算机测试证明该算法可以实现mat - u和/或反应器导流器功率平衡的实时控制。

The double-null divertor is an attractive exhaust solution for compact tokamak reactors, but its success is contingent on achieving vertical plasma position control which is both robust and precise, and uses some non-magnetic state estimation. A natural choice for this would be sensing the heat flux imbalance between the two divertors, since this quantity is exactly what we wish to minimise to ensure equal power sharing. The time delay between plasma moving vertically and the heat flux being detected - a part of which is accounted for by transience in the scrape-off layer (SOL) - is an important parameter in this control concept. If the information advects along the SOL quickly, then heat flux becomes a good estimator of the plasma displacement, but if it is slow (compared to the vertical oscillation timescale) then closed-loop control becomes much more challenging. It is unknown what this characteristic SOL time will be in a reactor, but it seems prudent to mitigate against the possibility it is 'slow' by devising an advanced vertical control scheme of the kind which would be required in this scenario. We propose to create a model predictive control algorithm which makes use of a fast reduced/surrogate model for the time-dependent SOL behaviour to predict, based on recent measurements, the future evolution of the system in real time and actuate on the vertical control coils accordingly. This model should be validated against MAST-U data and simulations, and the algorithm tested in-silico to demonstrate that real-time control of MAST-U and/or a reactor divertor power balance could be achieved by this method.