Real-time Plasma Control System Research Articles

Reconstruction of tokamak plasma safety factor profile using deep learning

The motional Stark effect (MSE) diagnostic has been a standard measurement for the magnetic field line pitch angle in tokamaks that are equipped with neutral beams. However, the MSE data are not always available due to experimental constraints, especially in future devices without neutral beams. Here we develop a deep-learning based model (SGTC-QR) that can reconstruct the safety factor profile without the MSE diagnostic to mimic the traditional equilibrium reconstruction with the MSE constraint. The model demonstrates promising performance, and the sub-millisecond inference time is compatible with the real-time plasma control system.

Real-Time Magnetic Sensor Anomaly Detection Using Autoencoder Neural Networks on the DIII-D Tokamak

Magnetic diagnostics in tokamaks are key to plasma equilibrium control (plasma current, plasma shape, and position) and amelioration of plasma instabilities. Thus, real-time identification of the anomalous sensor is mandatory. A new real-time system based on autoencoder (AE) neural networks (NNs) for anomaly detection in magnetics signals, including both flux loops and magnetic probes, has been successfully implemented on the plasma control system (PCS) of the DIII-D tokamak. The AE NN is trained on over 4000 plasma discharges, with an optimized latent space representation of the input signals while minimizing the reconstruction loss. An algorithm determining anomalous sensors based on excessive deviations from accurate reconstruction from the trained NNs is constructed in a MATLAB/Simulink environment and is deployed on the real-time PCS using the embedded MATLAB coder (EMC) environment. The reconstruction performance of the algorithm is quantified in hardware-in-the-loop (HIL) simulation by injecting artificially common fault types, including stuck-at-zero fault, saturated fault, noise fault, and drift fault. The AE NN algorithm has been successfully commissioned on the DIII-D PCS.

Deep learning based surrogate models for first-principles global simulations of fusion plasmas

The accurate identification and control of plasma instabilities is important for successful fusion experiments. First-principle simulations that can provide physics-based instability information such as the mode structure are generally not fast enough for real-time applications. In this work, a workflow has been presented to develop deep-learning based surrogate models for the first-principle simulations using the gyrokinetic toroidal code (GTC). The trained surrogate models of GTC (SGTC) can be used as physics-based fast instability simulators that run on the order of milliseconds, which fits the requirement of the real-time plasma control system. We demonstrate the feasibility of this workflow by first creating a big database from GTC systematic linear global electromagnetic simulations of the current-driven kink instabilities in DIII-D plasmas, and then developing SGTC linear internal kink instability simulators through supervised training. SGTC linear internal kink simulators demonstrate predictive capabilities for the mode instability properties including the growth rate and mode structure.

Preliminary design of real-time plasma control system for CFETR

The China Fusion Engineering Test Reactor (CFETR), aiming at demonstrating fusion energy production in self-sufficient burning in different advanced operation scenarios, has entered the stage of engineering design and R&D (research and development). The real-time plasma control system (PCS), as a critical system controlling the evolution of plasma parameters that are necessary to operate CFETR, is designed synchronously. The CFETR PCS is composed of system architecture, integrated control algorithms, and two assistant tools PCS-SDP (Software Development Platform) and PCS-VP (Verification Platform) for algorithm development and verification. Taking the steady-state operation and machine safety into account, the requirement and conceptual design of the system architecture of CFETR PCS is introduced in this paper. PCS-SDP, a visual algorithm development platform, has completed the preliminary design and a prototype has been built to validate the design and technical feasibility with the control algorithm development for EAST. PCS-VP is under preliminary design taking ITER PCSSP (Plasma Simulation Platform) design as reference. These two tool platforms will strongly support and accelerate the development of CFETR PCS.

Conceptual design of the COMPASS-U control systems

The COMPASS-U tokamak is an advanced and more complicated successor of the COMPASS tokamak. Almost all main technological parts will be different. Both the higher performance of COMPASS-U and the increased number of parts and systems lead to a higher risk of failure. Therefore, new control systems must be developed regarding handling dangerous situations and integrating new systems. The COMPASS-U also opens the opportunity to make significant changes in the current systems based on the experience gained from COMPASS operation and to upgrade the systems to nowadays technologies.This article focuses on the conceptual design of the COMPASS-U control systems. It shows all CODAC (Control, Data Access, and Communication) systems their purpose, functionality, operation principle, interaction, and main used technologies.The COMPASS-U will use EPICS (Experimental Physics and Industrial Control System) for Slow control and supervision system. The real-time plasma control system will utilize MARTe2 (Multi-threaded Application Real-Time executor). The base control loop period of the Real-time plasma control system will be 50 μs as of COMPASS.

Machine learning control for disruption and tearing mode avoidance

Real-time feedback control based on machine learning algorithms (MLA) was successfully developed and tested on DIII-D plasmas to avoid tearing modes and disruptions while maximizing the plasma performance, which is measured by normalized plasma beta. The control uses MLAs that were trained with ensemble learning methods using only the data available to the real-time Plasma Control System (PCS) from several thousand DIII-D discharges. A “tearability” metric that quantifies the likelihood of the onset of 2/1 tearing modes in a given time window, and a “disruptivity” metric that quantifies the likelihood of the onset of plasma disruptions were first tested off-line and then implemented on the PCS. A real-time control system based on these MLAs was successfully tested on DIII-D discharges, using feedback algorithms to maximize βN while avoiding tearing modes and to dynamically adjust ramp down to avoid high-current disruptions in ramp down.

Prediction of mechanical behavior of the KSTAR CS magnet based on PF coil currents

The KSTAR includes seven pairs of ring-shaped poloidal field (PF) superconducting coils with rectangular cross section. The central solenoid (CS) is a vertical stack of four pairs of PF coils compressed axially by preloading structures. The axial compression (remaining preload) on the CS coils is monitored by strain measurement, one of the important monitoring parameters for safe operation of KSTAR. The preload variation during plasma discharge can be predicted based on the PF coils and plasma currents without complicated structural analysis. The calculation of equivalent vertical EM force was proposed to estimate the bottom displacement of the CS magnet with the assistance of the joint diagram for bolted joints. The predicted vertical displacements according to the proposed methodology were reasonably well in agreement with the displacement measured during long pulse discharge #18437. The predicted displacement was used to quickly calculate the preloading reduction of the CS magnet structure. This approach will contribute to the development of operation scenarios for long pulse and large plasma current discharges. During plasma discharge, our algorithm can be applied to a real-time plasma control system.

Implementing a finite-state off-normal and fault response system for disruption avoidance in tokamaks

A finite-state off-normal and fault response (ONFR) system is presented that provides the supervisory logic for comprehensive disruption avoidance and machine protection in tokamaks. Robust event handling is critical for ITER and future large tokamaks, where plasma parameters will necessarily approach stability limits and many systems will operate near their engineering limits. Events can be classified as off-normal plasmas events, e.g. neoclassical tearing modes or vertical displacements events, or faults, e.g. coil power supply failures. The ONFR system presented provides four critical features of a robust event handling system: sequential responses to cascading events, event recovery, simultaneous handling of multiple events and actuator prioritization. The finite-state logic is implemented in Matlab®/Stateflow® to allow rapid development and testing in an easily understood graphical format before automated export to the real-time plasma control system code. Experimental demonstrations of the ONFR algorithm on the DIII-D and KSTAR tokamaks are presented. In the most complex demonstration, the ONFR algorithm asynchronously applies ‘catch and subdue’ electron cyclotron current drive (ECCD) injection scheme to suppress a virulent 2/1 neoclassical tearing mode, subsequently shuts down ECCD for machine protection when the plasma becomes over-dense, and enables rotating 3D field entrainment of the ensuing locked mode to allow a safe rampdown, all in the same discharge without user intervention. When multiple ONFR states are active simultaneously and requesting the same actuator (e.g. neutral beam injection or gyrotrons), actuator prioritization is accomplished by sorting the pre-assigned priority values of each active ONFR state and giving complete control of the actuator to the state with highest priority. This early experience makes evident that additional research is required to develop an improved actuator sharing protocol, as well as a methodology to minimize the number and topological complexity of states as the finite-state ONFR system is scaled to a large, highly constrained device like ITER.

Towards the conceptual design of the ITER real-time plasma control system

ITER will be the world's largest magnetic confinement tokamak fusion device and is currently under construction in southern France. The ITER Plasma Control System (PCS) is a fundamental component of the ITER Control, Data Access and Communication system (CODAC). It will control the evolution of all plasma parameters that are necessary to operate ITER throughout all phases of the discharge. The design and implementation of the PCS poses a number of unique challenges. The timescales of phenomena to be controlled spans three orders of magnitude, ranging from a few milliseconds to seconds. Novel control schemes, which have not been implemented at present-day machines need to be developed, and control schemes that are only done as demonstration experiments today will have to become routine. In addition, advances in computing technology and available physics models make the implementation of real-time or faster-than-real-time predictive calculations to forecast and subsequently to avoid disruptions or undesired plasma regimes feasible. This requires the PCS design to be adaptable in real-time to the results of these forecasting algorithms. A further novel feature is a sophisticated event handling system, which provides a means to deal with plasma related events (such as MHD instabilities or L-H transitions) or component failure. Finally, the schedule for design and implementation poses another challenge. The beginning of ITER operation will be in late 2020, but the conceptual design activity of the PCS has already commenced as required by the on-going development of diagnostics and actuators in the domestic agencies and the need for integration and testing. This activity is presently underway as a collaboration of international experts and the results will be published as a subsequent publication. In this paper, an overview about the main areas of intervention of the plasma control system will be given as well as a summary of the interfaces and the integration into ITER CODAC (networks, other applications, etc.). The limited amount of commissioning time foreseen for plasma control will make extensive testing and validation necessary. This should be done in an environment that is as close to the PCS version running the machine as possible. Furthermore, the integration with an Integrated Modeling Framework will lead to a versatile tool that can also be employed for pulse validation, control system development and testing as well as the development and validation of physics models. An overview of the requirements and possible structure of such an environment will also be presented.

Upgrade of the COMPASS tokamak real-time control system

The COMPASS plasma control system is based on the MARTe real-time framework. Thanks to MARTe modularity and flexibility new algorithms have been developed for plasma diagnostic (plasma position calculation), control (shaping field control), and protection systems (central solenoid protection). Moreover, the MARTe framework itself was modified to broaden the communication capabilities via Aurora.This paper presents the recent upgrades and improvements made to the COMPASS real-time plasma control system, focusing on the issues related to precision of the real-time calculations, and discussing the improvements in terms of discharge parameters and stability. In particular, the new real-time system has given the possibility to analyze and to minimize the transport delays of each control loop.

Plasma control system for “Day-One” operation of KSTAR tokamak

A complete plasma control system (PCS) has been developed for KSTAR’s first plasma campaign as a collaborative project with the DIII-D team. The KSTAR real time plasma control system is based on a conceptual design by Jhang and Choi [Hogun Jhang, I.S. Choi, Fusion Engineering and Design 73 (2005) 35–49] and consists of a fast real-time computer/communication cluster and software derived from the GA-PCS [Penaflor, B.G., et.al., Fusion Engineering and Design, 83 (2) (2008) 176]. The system has been used for simulation testing, poloidal field (PF) coil power supply commissioning and first plasma control. The seven sets of up-down symmetric, superconducting PF coil/power supply systems have been successfully tested. Reflective memory (RFM) is utilized as the primary actuator/PCS real-time communication layer and PCS synchronization with KSTAR timing system and slower control devices is achieved through an EPICS implementation. Consistent feedback loop times of 100 microseconds has been achieved during PF coil power supply testing and first plasma commissioning. Here we present the “Day-One” plasma control system in its final form for the first plasma experimental campaign of KSTAR and describe how the system has been utilized during magnet commissioning and plasma startup experiments.

Digital real-time plasma control system for Alcator C-Mod

A new digital plasma control system (DPCS) has been implemented for Alcator C-Mod. The new system was put into service at the start of the 2005 run campaign and has been effectively controlling the machine since then. This system not only provides a drop-in replacement for the original C-Mod control system, which was becoming increasingly difficult to maintain, it also provides a flexible platform to explore new non-linear and adaptive control schemes. It consists of two 64 input, 16 output CompactPCI digitizers attached to a rack mounted Linux server, which performs both the I/O and the computation. During initial operation, the system is set up to directly emulate the original C-Mod control system: this comprises a matrix of linearized observers, which estimate the 16 physical quantities under control, 16 PID channels for the conditioning of the errors between the physical quantities and their pre-programmed target waveforms and a matrix of controllers, for feeding back to the power supplies and other actuators. The matrices, the target and feed-forward waveforms and the other control parameters are time dependant and switch according to the particular phase of the plasma discharge. The compatibility with the previous control system allows the existing user interface software and data structures to be used with the new hardware. The controlling program is written in IDL, from Research Systems Incorporated (RSI), and runs under standard Linux. Interrupts are disabled during the plasma pulses to achieve real-time operation. Emulation of the original control algorithms is accomplished using 50 μs per iteration, with the time evenly split between I/O and computation, so rates of about 20 kHz are achievable. Reliable vertical control can be maintained with cycle rates as low as 5 kHz so there is considerable headroom for extra computation. The next steps are to explore the operation of the system on multiple time-scales, the use of additional computers in the real-time loop and the implementation of user provided signal processing and control schemes. These would include real-time analysis of power supply demands, real-time MHD modelling, and adaptive response to non-standard (unplanned) discharge conditions.

DIII-D integrated plasma control tools applied to next generation tokamaks

A complete software suite for integrated tokamak plasma control has been developed within the DIII-D program. The suite consists of software for real-time control of all aspects of the plasma, modeling, simulation and design tools for analysis and development of controllers, a flexible and modular architecture for implementation and testing of algorithms and many fully validated models. Many elements of the system have been applied to and implemented on NSTX and MAST. The DIII-D realtime plasma control system together with the integrated modeling and simulation suite have been selected for operational use by both the KSTAR and EAST tokamaks, and are also being used at General Atomics to investigate control issues for ITER.

Real-time identification of the resistive-wall-mode in DIII-D with Kalman filter ELM discrimination

Abstract The resistive-wall-mode (RWM) is a major performance-limiting instability in present-day tokamaks. Active control and stabilization of the mode will almost certainly be essential for the success of advanced tokamaks and for the economic viability of tokamak fusion reactors. High performance tokamak plasmas often experience edge-localized-modes (ELMs) which can interfere with RWM identification and control. If the RWM control scheme reacts to an ELM the RWM may be driven unstable instead of controlled. An algorithm for real-time identification of the RWM with discrimination of ELMs in the DIII-D tokamak has been developed using a combination of matched filter and Kalman filter methods. The algorithm has been implemented in DIII-D’s real-time plasma control system (PCS) and is available to drive active mode control schemes.

JT-60 Control System

The present status of the JT-60U control system is reported including its original design concept, the progress of the system, and various modifications since the JT-60 upgrade. This control system has features of a functionally distributed and hierarchical structure, using CAMAC interfaces initially, which have been replaced by versatile module Europe (VME)-bus interfaces, and a protective interlock system composed of both software and hard-wired interlock logics. Plant monitoring and control are performed by efficient data communication through CAMAC highways and Ethernet with TCP/IP protocols. Sequential control of plasma discharges is executed by a combination of a remodeled VME-bus system and a timing system. A real-time plasma control system and a human interface system have been continuously modified corresponding to the progress of JT-60U experiments.

Real-time disruption handling at ASDEX upgrade

A neural network for prediction of disruptions has been developed at ASDEX Upgrade with the goal to mitigate or avoid these. The novel idea is to compute the remaining time-to-disruption to indicate the stability level of the discharge. The neural network has been specified, trained and then implemented within the real-time plasma control system. The current version of the system terminates the discharge with an impurity pellet when the computed time-to-disruption falls below a threshold of 80 ms. Routine operation shows that disruptions are recognized reliably. Vessel currents and forces are considerably reduced. The system will be enhanced to avoid disruptions with a soft landing initiated in time.

The MAST data acquisition system — system architecture

A new data acquisition (DA) system has been developed for the MAST experiment at Culham. It has also been implemented on the COMPASS experiment as a replacement for the old PDP-11 based data acquisition system and as a test bed for MAST. The DA system is distributed across a number of computers with a DA process for each distinct diagnostic system. An object-orientated approach is taken to the control and readout of each device in the system. It is designed to be independent of the hardware interfaces used on each diagnostic. The system is flexible enough to cope with diagnostics ranging from those involving simple time evolving signals to complex spectrometers, and will incorporate a new high speed distributed timing system. This system is also being considered as the interface to the real time Plasma Control system on MAST. A distributed scheduling system is used to co-ordinate the activity of each DA process with the Central Control system for each experiment. This paper describes the architecture of this data acquisition system with particular emphasis on the core of the data acquisition system. Aspects of distributed implementation of the system on real diagnostics are discussed in a companion paper [1].

The real time plasma control system of Tore Supra

The Tore-Supra real time plasma control system, operational since 1988, was completely redesigned in 1995 with VME hardware and PowerPC processors monitored by a real time UNIX operating system (LynxOS™). Now, with five units interconnected through a fast shared memory network, a global plasma control has been built. Plasma current, shape, and gas control are already established and it is expected that control of the plasma current profile will be achieved before the end of 1997. The introduction of a specific unit with a graphic man-machine interface in the shared memory network is under development which will allow interactive verification of parameters. This new architecture is flexible and open-ended. It will enable interconnection to other Tore Supra systems such as those required for the long-term programme of long pulse (1000s) discharges.

Real-time plasma control system for LHD

The large helical device (LHD) is the largest superconducting heliotron-type device under construction at the National Institute for Fusion Science (NIFS) at Toki, Japan. One of the key issues of the LHD project is the demonstration of currentless steady-state plasma operation. To achieve steady-state operation, real-time plasma control (RTPC) is designed. The feedback control system for plasma current is made for the test of the time response and the algorithm. The fuzzy algorithm has been compared with the standard PID algorithm. The fuzzy control shows better performance for the nonlinear control response than the standard PID control.

Real time data acquisition system for control and long pulse operation in Tore Supra

The Tore Supra Data Acquisition System was completely redesigned in 1996, for plasma control and long pulse operation. It is now based on the Rtworks package which provides all the software basic modules to build a distributed on-line system, from data acquisition level up to run control and data display. In the same time, the real-time plasma control system has been improved with several control units upgraded with VME hardware and PowerPC processors interconnected through a fast shared memory network.