Dynamic pulse scheduling in ASDEX Upgrade: Disruption avoidance and investigation of the H-Mode density limit

Operation of a tokamak device requires the coordinated operation of a multitude of systems. The sequence of operations during a discharge is both too complex and fast for human interaction. A common way of operation is a predefined sequence of operations which utilizes continuous control to achieve the desired discharge. This however poses the risk of e.g. disruptions, if an unforeseen event occurs which would require a different action than the programmed one. Furthermore, in case of larger devices, e.g. ITER, major disruptions are not acceptable and exception handling is required which can react to unplanned events. The ASDEX Upgrade discharge control system (DCS) is able to detect events and change the control goals dynamically, applying complex continuous control with multiple sensors and actuators. This contribution discusses how these capabilities are utilized for disruption avoidance and the study of the H-Mode density limit (HDL). It is demonstrated that disruption avoidance for the HDL is possible using event based exception handling. For the detection of the HDL both direct measurements of the occurring MARFE and a state space model representing the HDL in terms of multiple measurable quantities have been used.Avoiding the disruption in case of the HDL extends the stable operation space of ASDEX Upgrade. This has been utilized to perform multiple automated experiments within one discharge which significantly decreased the number of discharges required for the physics investigation of the HDL.The combination of exception handling and continuous control has enabled studies of the HDL, in which the control system on its own sets up the desired target scenario. Extending these capabilities enables the automatic scanning of the accessible operational space of a device without the need of a scheduled disruption. This would be especially useful for devices where the number of allowable disruptions is limited and the acceptable disruption risk needs to be assessed before the experiment.