Fusion Engineering and Design 73 (2005) 35–49
Design concepts for KSTAR plasma control system Hogun Jhang ∗ , I.S. Choi Korea Basic Science Institute, 52 Yeoeun-dong, Yusung-Gu, Taejon, Daejeon, Republic of Korea Received 24 August 2004; received in revised form 24 December 2004; accepted 28 December 2004 Available online 9 April 2005
Abstract We present design concepts and features of the plasma control system (PCS) in Korea Superconducting Tokamak Advanced Research (KSTAR). A design structure of the PCS is proposed as an effort to achieve research objectives of the KSTAR project. The PCS architecture is characterized by the real-time data communication using reflective-memory network, the hierarchical organization of dedicated controllers, and the integrated generation of actuator signals. Discussions are made of the functions and present design choices of the KSTAR PCS. © 2005 Elsevier B.V. All rights reserved. Keywords: Plasma control; System design; KSTAR
1. Introduction The research objectives of the proposed Korea Superconducting Tokamak Advanced Research (KSTAR) project [1] are the extensions of present operational boundaries of tokamak plasmas through the active control of magnetohydrodynamic (MHD) activities and transport, and the realization of steady state advanced tokamak (AT) operations by the exploitation of noninductive current drives. Major target plasma parameters and design features of the KSTAR device were discussed in Ref. [1], and are summarized in Table 1. The research objectives of the KSTAR project call for the development of a reliable and powerful plasma control ∗ Corresponding author. Tel.: +82 42 865 3607; fax: +82 42 865 3469. E-mail address:
[email protected] (H. Jhang).
0920-3796/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.fusengdes.2004.12.004
system (PCS). Especially, it is required to control not only the usual equilibrium parameters, such as plasma position, shape, and current, but also MHD activities and kinetic parameters, since the AT experiments are to be carried out near operational boundaries (or even passing through them) imposed by MHD activities. We define the KSTAR PCS as a system of hardware and software to identify plasma parameters and/or plasma operation status, to evaluate necessary actuator signals for the correction of errors, and to execute synchronized action in order to achieve a discharge goal. From a viewpoint of the KSTAR research program, the PCS plays the important role of integrating developments of AT operation scenarios, real-time plasma control diagnostics, and advanced actuators. From a practical point of view, however, designing an advanced PCS is a big challenge because of limited knowledge available. There are uncertainties to be re-
36
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
Table 1 Major target parameters of the KSTAR device Parameters
Baseline
Upgrade
Toroidal field, BT (T) Plasma current, Ip (MA) Major radius, R0 (m) Minor radius, a (m) Elongation (κx ) Triangularity (δx ) Poloidal divertor null Safety factor, q95 Plasma volume (m3 ) Plasma surface area (m2 ) Pulse length (s)
3.5 2 1.8 0.5 2 0.8 2 3–10 ∼16 ∼48 20
300
Heating power (MW) Neutral beam Ion cyclotron Lower hybrid Electron cyclotron
16 8 6 1.5 0.5
22 14 6 1.5 0.5
Annual deuterium operation time (s) Number of pulse
20000 50000
solved for the effectuation of an advanced PCS design. First, the physics understanding of dynamic responses of tokamak plasmas under plasma control actions is still premature, except for some cases of equilibrium control whose controller model has been well established for decades [2–7]. It can bring about the failure of a control system by misinterpretating correlated plasma parameters; some control action might deteriorate other plasma control parameters. Second, the controllability of a large number of control parameters with relatively limited number and capability of actuators has not been fully assessed. Further, the feasibility of exploiting advanced control algorithms has not been well demonstrated, although some efforts have been reported in the literature for plasma equilibrium control [5,8]. Third, the KSTAR device consists of a large number of subsystems being operated in distinct hardware and software environment. The PCS itself is also divided into fundamental units running distinct control algorithms. The large number of heterogeneous tokamak subsystems and plasma control units should be combined in a cohesive manner to ensure a reliable and safe tokamak operation. In these respects, we have established the system flexibility as the design principle for the KSTAR PCS. The PCS should be designed to be capable of providing an easy upgrade, whenever requested. A
Extended option
≤27.5
ECCD
great deal of configurational and operational flexibility is necessary both in hardware and software levels. The hardware chosen should be independent of the software to be installed. Care must be taken to establish the PCS architecture even in conceptual level to guarantee the system flexibility. In order to accomplish a flexible control system, the PCS should be designed to provide a high degree of modularity in such a way that frequent software changes do not affect other interfacing systems. The clear and simple interface definitions between related systems can materialize the modular and autonomous structure, facilitating the self-evolution of the PCS. In addition to the system modularity, the PCS should be compatible with the hardware and software environment of the overall KSTAR control and data system. It is because of the fact that the KSTAR PCS is regarded as one of the subcomponents constituting the whole KSTAR device. The compatibility with other relevant systems can accommodate the PCS to the whole KSTAR control system environment with minor efforts. In this paper, we describe design concepts and features for the KSTAR PCS under the guideline of the aforementioned design principle and requirements. Section 2 starts with the design overview of the whole KSTAR control and data system and its relationship with the PCS. Design feature of the KSTAR PCS ar-
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
37
Fig. 1. An overview of the KSTAR control and data system (KCODA) and the integration of the plasma control system into the KCODA.
chitecture are also presented in Section 2. Sections 3 and 4 are devoted to brief discussions on functions of operation control and magnetics control, respectively. The present design status and some issues are also described in Section 4. This paper concludes with a brief summary in Section 5.
2. Design architecture of the KSTAR PCS 2.1. Design overview Fig. 1 shows the schematic diagram of the overall KSTAR control and data-acquisition system (KCODA) and the integration of the PCS into the KCODA. The supervisory control system (SCS) is responsible for the integration of facility control, machine control, data-acquisition system, timing system, and the plasma control system. Four classes of PCS subsystems are indicated by grey boxes in Fig. 1. The operational control takes the charge of discharge preparation and management during a plasma discharge. Dedicated plasma control systems consist of the other three plasma con-
trol classes, the magnetics control, the kinetics control and the disruption alarm and fast shutdown systems. Each of the dedicated control systems consists of several elementary control units (ECUs) the detailed description of which will be presented in Section 2.4. Fig. 2 depicts an overview of software environment for the KCODA. Experimental Physics and Industrial Control System (EPICS) [9], which has been widely used in large facilities, especially in the accelerator community [9], has been selected as the main development tool for the KCODA. In tokamak experiments, only the NSTX control system has been developed using EPICS complemented by the Model Data System Plus (MDS+) [10] for data-acquisition [11]. Unlike NSTX, however, we will use MDS+ as an alternative database system only for after-shot analyses by exporting data to a MDS+ tree structure. In this way, we can provide an easy access to physicists who are familiar with the MDS+ system. The EPICS consists of a core and a set of extensions that provide a number of tools for creating a control system, minimizing the need for custom coding and ensuring uniform operator interfaces. EPICS channel ac-
38
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
Fig. 2. The general software architecture of the KSTAR control and data system.
cess is based on the client–server model, and provides network transparent access to input output controller (IOC) databases. All of the software applications are connected to the EPICS control system by the channel access server (CAS) through which the dedicated real-time plasma control units communicate with the SCS. The implementation concept of the PCS within
the EPICS environment will be described in Section 2.2. 2.2. The PCS architecture The conceptual design features of the KSTAR PCS in order to materialize the design principle can be sum-
Fig. 3. A conventional form of the plasma control system.
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
marized as: (1) the employment of a fully digital control system, (2) the real-time data communication through reflective memory, and (3) the compatibility with the EPICS software environment. Fig. 3 shows a typical example of a conventional feedback control system that has been employed in present tokamak experiments. The supervisory control system prepares for a discharge by setting up reference values of control parameters and controller information and transferring them to relevant plasma controllers prior to the discharge. Usually, the communication between the supervisory system and the embedded plasma control system is closed. The reference values are not changed during the plasma experiment. A dedicated controller receives measurement data directly from the relevant diagnostic devices, evaluate errors in control parameters by the interpretation of diagnostic signals, calculate actuator signals to correct the errors, and access to corresponding actuators. This conventional form of a feedback controller has been used successfully for the equilibrium control in present toka-
39
maks [3,12–15]. It is a simple but efficient approach, when the relevant diagnostics are manifest, involved controllers are independent each other, and the plasma response model (plant model) is well established. In the case of plasma kinetic parameter control, which is envisioned to be essential for AT operations, the causal relationship between the relevant diagnostics and plasma controllers are often ambiguous due to the lack of plasma response model. Since the dedicated controllers are no longer supposed to be independent in this case, diagnostics, controllers, and actuators form a complicated control matrix [16] or control network. Then, it is likely that a diagnostic signal is related to many controllers, and different controllers are linked to one actuator. The structural and configurational flexibility is likely to be reduced if we design a PCS by adopting the conventional approach. In order to resolve this problem, at least in the conceptual level, the KSTAR PCS adopt the control system architecture shown in Fig. 4. The essential points of Fig. 4 are that there are: (1) a common communication layer
Fig. 4. The architecture of the KSTAR plasma control system.
40
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
Fig. 5. Implementation concept of the KSTAR plasma control system into the EPICS environment.
provided by a real-time shared memory network through which control data are shared by the entire system in real-time and (2) an arbitrator through which the actuator signals are generated by some arbitration algorithms the concepts of which will be discussed in Sections 2.3 and 2.4, respectively. Since the software environment of KCODA will be based on EPICS, the implementation scheme of the PCS architecture should be conceptualized within the EPICS environment. Fig. 5 represents a conceptual scheme for the implementation of the KSTAR PCS in the EPICS environment. Data that are not time-critical communicate through the LAN via EPICS CAS. In the development stage of the KSTAR control system, the EPICS-shared memory device/driver support has been developed to provide the access to the real-time shared memory network. The driver modules from the relevant diagnostics and actuators to the shared memory should also be provided, which are under development. 2.3. Real-time data communication In KSTAR, the real-time data communication among processed diagnostic signals, the SCS, and dedicated controllers are accomplished through the common communication layer. The exploitation of the real-time common communication layer for a tokamak
plasma control system has been developed by Raupp et al. [17] to integrate heterogeneous systems. It is under implementation process at the ASDEX-U device [18]. The merits of this approach are that it can ensure the easy integration of heterogeneous control hardware and software involved in various control components, the system flexibility, and the interconnection of distinct control units, at least in the structural level despite of the lack of controller model. As a hardware solution for the common communication layer, we have chosen the real-time network with a ring topology based on the VMIC fiber-optic reflective memory. The selected hardware for the prototype KSTAR PCS is the PCI bus mezzanine card (PMC) version of the reflective memory, VMIPMC-5565 [19]. It provides the 2.1 Gbaud speed with a 174 MB/s bandwidth and the node latency of 40 ns. Since the maximum real-time data flow rate in KSTAR will not exceed 8 MB/s (∼2000 words/ms) even in its full operation, the reflective-memory network amply satisfies the bandwidth requirement of the KSTAR PCS. In addition to the reflective memory solution described above, the hardware system for the prototype KSTAR PCS is under development based on the Compact PCI standard and DSP processors [20]. The PCS hardware is controlled by the pentium-based Kontron CP6000 system controller to download onto the DSP
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
Fig. 6. The structure of a model elementary control unit X in the KSTAR plasma control system.
farm and to control the whole system. Linux Red Hat 9 operating system (OS) is used for the local controller to configure the system prior to a discharge. Details of the KSTAR prototype hardware system and the results of performance test will be reported in the future. 2.4. Generation of actuator signals Actuator signals are generated by the participating ECUs, the schematic view of which is illustrated in Fig. 6. Each ECU contains: (1) physics-oriented computational processes and (2) implementation-oriented protocol processes. The computational processes include the identification of specific plasma control parameters by the manipulation of diagnostic data, and the calculation of actuator signals by the use of some control algorithms. For instance, the process M SHP IDT = CDE is recognized as the process for the shape identification using the current element method [21]. An ECU accesses to the real-time reflective memory network to acquire relevant processed diagnostic data written by plasma control-related diagnostics. Also, it writes its current status onto the specified address in
41
the reflective-memory network, and receives the current discharge and plant status. An ECU has its own control cycle within which the controller action items must be completed. The base cycle is defined as the minimum control cycle among the ECUs. It was chosen to be 1 ms, which is required for the control of plasma position. However, the system is designed to be able to alter the base cycle (1, 2, 5, 10 ms) according to experimental conditions. A control cycle for an ECU will be a multiple of the base cycle, as shown in Fig. 6. An ECU sends acknowledgements messages to the cycle administrator (supervisory administration master), whenever it starts and finishes the control cycle to check any violation of the controller cycle [17]. A local administration master process supervises the violation of scheduled processes in the ECU. Also, it communicates with the SCS in real-time to receive information on the synchronization, scheduling and plant status. The main programming language for the controller development will be C and/or C++. Since each ECU receives reference values launched by physicists prior to a discharge, it possesses an onboard local memory to store the reference control values and control matrix coefficients required for the identification of control parameters and the computation of actuator signals. The actuator signals that are generated by participating ECUs are transferred to the arbitrator, as shown in Fig. 4. The arbitrator evaluates and issues final actuator signals using some arbitration algorithms. Thus, it hides the details of embedded controllers from the actuators. This approach, if successful, is expected to resolve the problem of conflicts and collisions between actuator signals issued by different control units, at least in a structural level. It provides an optimum choice of actuator signals to accomplish the discharge goal in the presence of a large number of control units, while the number of available actuators are limited. An arbitration algorithm should meet the discharge goal issued by the supervisor on the basis of as closely as possible principle. The arbitration algorithm should also include the limitation of actuators (the limitation of currents in superconducting poloidal field coils, for instance) to guarantee the machine protection. The simplest conceivable algorithm is to assign a weighting factor for each dedicated control unit by ranking the criticality of a particular control unit achieving the discharge goal. In this case, the weights in the ECUs can be determined
42
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
in real-time by the consideration of plasma discharge status. In the early phase of KSTAR operation, when the arbitration algorithm will have not been established, the arbitrator will be served as a monitor to detect any discrepancies between actuator signals requested by different ECUs. So far, no applicable arbitration algorithms have been developed beyond the conceptual level. We plan to conduct a research on the development of an arbitration algorithm, including the assessment of controller response in the presence of the arbitration algorithm.
zation flux of ψ0 = −6.1 Wb and poloidal field null condition. Here, ψ0 is the poloidal flux provided by the PF coils, linking through the geometric center of the KSTAR device (R = 1.8 m, Z = 0.0 m). The initial PF coil currents are calculated by solving the 2Nc +1 system of equation, NPF
j
j
Gij (R0 , Z0 ; RPF , ZPF )Ij = ψ0
j=1
NPF 1 ∂Gij Ij = 0, Ric ∂Z (Ric ,Zci ) j=1
3. Roles of operation control
NPF 1 ∂Gij Ij = 0. Ric ∂R (Ric ,Zci ) j=1
The operation control takes the responsibility of the discharge management during the plasma operation. Main roles of it are summarized as: (1) the experiment planning by providing a discharge scenario, (2) the discharge phase/status identification, and (3) the judgment of discharge termination or continuation. 3.1. Discharge preparation A plasma discharge is prepared by calculating the reference waveforms of plasma control parameters and actuator signals. Prior to a discharge, controller characteristics and the corresponding control strategy are also evaluated. Then, they are transferred to relevant ECUs and actuators. The conventional man machine interface (MMI) tools, such as IDL, MATLAB, and EPICS operator interface (OPI) are used in this process. Some plasma simulation tools, for instance the MGAMS algorithm that has been used for the plasma position and shape control in TCV [4], are also used for this purpose. The same information is also delivered to the MDS+ database, and either MDS+ or EPICS loader will transfer reference values to the relevant subsystems. 3.2. Discharge execution Fig. 7 depicts an example of KSTAR plasma discharges. A plasma discharge is divided into several phases according to the physics goals to be achieved. (1) Initial magnetization: The poloidal field (PF) coils are charged up in order to provide initial magneti-
Here, Nc is the total number of constraint points satisfying the poloidal field null condition, NPF is the total number of independent PF coil circuits, Ij is the current of the jth PF coil, (R0 , Z0 ), (Ric , Zci ), j j and (RPF , ZPF ) is the cylindrical coordinate for the geometric center, the ith field null constraint point, and the jth PF coil position, respectively, and Gij is the Green function between them. The singular value decomposition (SVD) technique [22] has been employed to solve this problem. When the measured PF coil currents agree with the preprogrammed values, the operation control proceeds the next phase. The duration of initial magnetization phase is 30 s. (2) Wait: After the initial magnetization phase, the 10 s of wait phase follows to cool down the heat generated in PF coils during the initial magnetization phase. (3) Plasma initiation: The plasma initiation phase follows the wait phase. The plasma breakdown occurs and a circular plasma forms with a minor radius of about 30 cm and a plasma current of 0.1 MA. The electron cyclotron heating of 500 kW is applied to assist a reliable plasma breakdown. For a reliable plasma initiation, a magnetic field null with sufficient quality, size and duration must exist within the vacuum vessel when a loop voltage is sufficient enough to generate avalanche breakdown. In KSTAR, the reference loop voltage for the inner start-up (Rstrt = 1.6 m) is 5 V. A conventional
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
43
Fig. 7. A typical plasma discharge scenario for KSTAR experiments.
measure to obtain a reliable Ohmic breakdown is known as JET formula, EBT /Bp > 1 kV/m, where E is the electric field at the breakdown position and Bp is the stray field generated inside the field null region. Then, the maximum allowable stray field for Ohmic breakdown in KSTAR is estimated to be Bp ≈ 5 (V) × 3.5 (T)/2π × 1.6 (m) (mT) ≈ 2 (mT). Since the eddy currents flowing nearby passive structures are expected to be the main source of the stray fields, it is of importance to assess the stray fields generated by eddy currents during the breakdown period. Time evolutions of stray fields during a 5 V inner start-up simulation are shown in Fig. 8. The maximum stray field near the field null position is found to exceed 10 mT, which is intolerable to guarantee a reliable Ohmic breakdown. Thus, they should be corrected by either the adjustment of PF coil currents at the initial magnetization phase or an active control of field null quality using in-vessel control coils, the study of which is under investigation. (4) Current ramp-up: After the plasma initiation phase, the plasma current is ramped up from 0.1 to 2 MA. The plasma minor radius increases and plasma shape evolves from a circular to a diverted
D-shape. Not only an accurate preprogramming of PF coil currents, but also an active control of a plasma current and shape are necessary to obtain correct plasma current and shape evolution. In the case of a standard Ohmic ramp-up, the duration of current ramp-up phase is 4 s, and ∼12 Wb of Volt-s is necessary to achieve a full plasma current (2 MA). An alternative approach, the non-inductive current ramp-up scenario, is worth investigating to mitigate the burden on the
Fig. 8. Time histories of stray fields generated by eddy currents during a 5 V inner start-up simulation.
44
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
superconducting PF coil system, hence reducing the quench risk of PF coils, by providing some portions of the full plasma current by means of non-inductive current drives [23]. (5) Heating and flattop: Auxiliary heating and current drive systems are applied to increase the plasma beta and to achieve the discharge goal. In KSTAR, major heating and current drive systems are 8 MW NBI, 6 MW ICRF, and 1.5 MW LHCD, as shown in Table 1, and corresponding real-time actuator signals comprise of NBI heating power, power and phase angles of rf systems, and voltages in piezoelectric valves in gas-feeder systems. Most of MHD and kinetics control activities are required in this phase to achieve the AT operation. (6) Current ramp-down: It is necessary to ramp down not only the plasma current but also the plasma density and external power input during this phase. Any disruptions caused by the excess of density or beta limits should be avoided. 3.3. Off-normal event handling When the abnormal behavior of machines (plant) or the plasma discharge is detected, the PCS terminates the discharge or changes the discharge status according to the prescribed action list. The origin of offnormal events is classified as the human safety-related issues, the machine and/or facility system faults, the plasma disruption or operation sequence violation, and the malfunction of PCS units (including the SCS itself). The discharge termination procedure will be decided based on the following criteria: (1) Level X, immediate shutdown of the discharge: - human safety-related events decided and issued by the SCS; - critical machine and/or facility interlock decided and issued by local hosts; - unavoidable plasma disruption issued by the disruption alarm process; - critical malfunction of the PCS decided and issued by the SCS. (2) Level Y, coordinated shutdown of the discharge: - machine and/or facility protection decided and issued by local hosts; - unavoidable plasma disruption precursor decided by the disruption alarm process;
- unrecoverable plasma discharge decided by the PCS; - non-critical malfunction of the PCS decided and issued by the SCS. (3) Level Z, watch out the subsequent status and change the discharge phase/status: - non-critical machine and/or facility malfunction decided and issued by local hosts; - plasma discharge phase/status violation. When the PCS detects a Level X event, it triggers the killer pellet-injector system to destroy the plasma in order to ameliorate the electromagnetic and mechanical loads induced by the disruption. At the same time, the SCS issues a command to discontinue the actions of all the actuator systems and dedicated ECUs. If a Level Y event occurs, the SCS commences the soft landing procedure, which is identical to the ramp-down phase in normal discharges. For a Level Z event, the PCS can either continue the discharge in watch-out status or start the soft landing procedure according to the severity of the event. After some post discharge processes, such as acquired data transfer, the machines and facilities are checked to investigate the origins of the off-normal event.
4. Functions of magnetics-control systems So far, most of PCS development activities in KSTAR have been focused on the plasma magneticscontrol studies because of its impact on the machine design. The magnetics control implies the feedback control of plasma equilibrium parameters by magnetic means. In KSTAR, it encompasses the fast time scale (1 ms cycle) plasma position control, the slow time scale (5–10 ms cycle) plasma current and shape control, the control of resistive wall modes (RWM) (10 ms cycle), and the compensation of non-axisymmetric error fields (continuous). The major measurement tools for the magnetics control are magnetic diagnostics, which are located inside the plasma facing components (PFC). They comprise of 45 flux loops, 48 tangential magnetic probes, 28 normal magnetic probes, and a system of Rogwskii coils [24]. The magnetic measurements for the control of RWM have not been settled yet, as will be discussed shortly. The superconduting PF coils are utilized for
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
45
Fig. 9. Axisymmetric structure model of the KSTAR device.
the control of plasma shape and current, while the segmented in-vessel control coils (IVCC) [25] are used for the control of plasma position, RWM, and field-error compensation (FEC). Fig. 9 represents the axisymmetric structure model of the PF coils, the passive conductors, and the location of IVCC coils. Fig. 10 illustrates the segmented IVCC design approach adopted in KSTAR. The IVCC system is segmented into 16 coils the poloidal location of which is given in Fig. 9 (four toroidal segmentation at each poloidal position). A significant engineering merit of this design approach is that each segmented coil can be inserted into the vacuum vessel through three NBI ports and one RF ports from outside of the vacuum vessel, as shown in Fig. 10(a). A careful connection method has been established in order that the segmented IVCC system can provide the position control, RWM control, and FEC simultaneously. Fig. 10(b) sketches the cross-sectional view and the connection scheme established for the IVCC system. Each coil is comprised of eight turns. The six turns of IVCU and IVCL given in Fig. 10(b) are serially connected to adjacent turns by
Fig. 10. (a) Configuration of KSTAR vacuum vessel and segmented IVCC in KSTAR. (b) Cross-sectional view of the segmented control coils and the connection scheme for magnetics control.
joining them externally to function on vertical position control, while the four turns of IRCU and IRCL are connected to provide radial position control. Two turns of IVCU (IVCL) and IRCU (IRCL) are connected to form the upper (lower) RWM/FEC coil, and the other two turns of IRCU and IRCL are connected to complete the middle RWM/FEC coil. We refer Ref. [25] for details of IVCC design and engineering analyses. 4.1. Plasma position control The control cycle of the vertical instability (1 ms) determines the base cycle of the KSTAR PCS. In spite of their high elongation (κx = 2.0), the growth rates of the vertical instability of KSTAR plasmas are significantly reduced due to the placement of close-fitting passive plates near the plasma boundary. The worst growth time has been found to be ∼10 ms for a low
46
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
beta, high li plasma in a full shape [26], where li is the internal inductance of the plasma. The segmented IVCC system shown in Fig. 10 is utilized as actuators for the position control. Based on Tokamak Simulation Code (TSC) [27] simulations [26], a proportional and derivative (PD) control [2] law and a simple proportional control logic has been chosen for the control of vertical and radial position, respectively. The power supply requirements for the vertical position control were also evaluated from TSC simulations. The maximum feedback current and voltage for the vertical position control has bee found to be 42 kA-turns and 123 kV/turn, respectively, from a stepresponse simulation [28]. For the radial position control, the required feedback current (voltage) has been evaluated from a position control simulation of a high beta plasma undergoing edge localized mode (ELM)like oscillations, yielding 22 kA-turns (52 V/turn). 4.2. Plasma shape and current control Avoidance of the contact of a plasma with PFCs during a steady state hot plasma operation requires the precise control of a plasma shape. The shape controller is a typical example of the multi-input multi-output (MIMO) control system. The full set of the KSTAR magnetic diagnostics will be used for the purpose of plasma shape identification, although it has been reported that a properly chosen reduced set could yield the similar identification accuracy [29]. Actuators for the plasma shape and current control are seven independent PF coil circuits (11 independent circuits for a single null operation). A model shape controller for KSTAR has been developed using a standard PID control law with optimized weighting factors for PF coils to reduce the total power consumption during the shape control action [29]. The required feedback current for each PF coil, Ij , is calculated by minimizing the function, 2 Nc NPF NPF Gij Ij − ψi + w2j Ij2 , ε= i=1
j=1
j=1
where Nc (NPF ) is the total number of shape control points (PF coil circuits), Gij is the Green function between the ith control point and the jth PF coil, ψi is the calculated flux error between the ith control point and
Fig. 11. (a) Time histories of shape errors for four shape-control points during a shape control simulation using TSC. A high β plasma (βN = 5.0) undergoing 10% permanent drop in stored energy has been used in the simulation. (b) Time evolution of currents in seven PF coils during the shape control simulation. Voltage limitation in each PF coil power supply has been taken into account.
a reference point (X-point for a diverted plasma), and wj is the weighting factor corresponding to the jth PF coil. Fig. 11(a and b) shows the time evolution of four shape control points (outermost flux surface (R0 + a, 0), innermost flux surface (R0 − a, 0), nearest point to passive plates (Rp , Zp ), and outer strike point (Rst , Zst )) and seven PF coil currents, respectively, during the shape control simulation. As a model disturbance for the simulation, a high beta plasma (βN = 5.0) undergoing 10% permanent β drop has been considered. An issue for the shape controller in KSTAR employing superconducting PF coils is to minimize the PF coil quench risk induced by the shape control action. A proper choice of PF coil weighting factors in real-time is likely to provide a possible remedy for the problem. Real-time recognition of PF coil states and an adjustment of weighting factors are expected to enable transfer of quench risk of the most vulnerable coil at
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
47
4.4. Field-error compensation
Fig. 12. Critical gain over which the n = 1 RWM is stable, as a function radial feedback coil position. The present position of the KSTAR IVCC is indicated.
the given time to other coils that are far from the superconduting stability boundary. The detailed study of it is under investigation and will be reported in the future. 4.3. Resistive wall mode control It has been known that the extension of the operational boundary beyond the no-wall limit demands for the stabilization of RWM [30]. In KSTAR, the FEC/RWM connection of the IVCC system given in Fig. 10(b) will be used for the active feedback stabilization of the RWM. The required IVCC feedback currents have been evaluated using an eigenvalue formulation in the cylindrical geometry by using the smart shellcontrol scheme [31] with a proportional control law. Fig. 12 shows the calculated value of critical gain over which the RWM is stable, as a function of feedback coil position. Using the present position of KSTAR IVCC system, it was shown that the RWM can be suppressed with proportional gain larger than 200. The required feedback current is estimated to be 2.4 kA-turns under the assumption of 5 G of RWM amplitude. There are remaining issues concerning about the design of the RWM control system. First, the selection and design of relevant magnetic diagnostics for the detection of RWM activities have not been completed yet. Second, the presence of poloidally partial passive plates in KSTAR (See Fig. 10) can also affect the optimization of the control algorithm. The optimization of the RWM controller in the presence of a poloidally partial structure is a non-trivial problem, which is a current R&D issue in KSTAR.
It is necessary to compensate the non-axisymmetric error fields introduced by the deviation from the perfect axisymmetry during the coil manufacturing and installation processes, in order to avoid disruptions induced by locked modes [32,33]. A recent finding of the effects of error field amplification on the growth of RWM supplements the indispensability of the error field compensation in future tokamaks [34]. In KSTAR, the FEC/RWM connection scheme of the IVCC system shown in Fig. 10(b) will be exploited for this purpose. A dc correction current is applied to the FEC/RWM coils for the entire period of a discharge. The power supply design has been based on the calculation of a dc current to correct field errors up to n = 2 mode (n, toroidal mode number) with 99.8 % confidence when σ = 2.0 mm, µr = 1.10, where σ is the standard deviation, and µr is the relative permeability of the KSTAR vacuum vessel [25]. The maximum required current has been estimated to be 13 kA-turns.
5. Summary and concluding remarks It will not be an overstatement that the development of a reliable and powerful plasma control system (PCS) for KSTAR is the most demanding prerequisite to fulfill the physics objectives of the KSTAR project. In this paper, we described the design concepts and features of the KSTAR PCS for the embodiment of the KSTAR research objectives given in Ref. [1]. The flexibility has been set as the design principle for the KSTAR PCS. Design features of the PCS architecture to materialize the design principle are the employment of real-time reflective-memory network to share critical information in real-time, and the exploitation of arbitrator for the integrated generation of actuator signals. The overall software architecture of the KSTAR control system will be based on the EPICS environment, and the implementation concept of the KSTAR PCS in EPICS environment has been established. Even though the proposed architecture seems to possess the potential to resolve many intricate problems of tokamak plasma control in the conceptual level, much R&D works are required to concretize the conceptual architecture of the KSTAR PCS.
48
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49
The KSTAR PCS is a subcomponent of the overall KSTAR control and data system, which is managed by the supervisory control system (SCS). The PCS is divided into four classes, operation control, magnetics control, kinetics control, and disruption alarm and fast shutdown, according to their functions. The operation control is in charge of discharge management during a plasma discharge. It prepare the discharge by generating reference values of plasma control parameters and feedback information, supervise the plasma discharge, monitor the punctuality of the current plasma phase with the preprogrammed sequence, and execute the plasma discharge in a coherent manner. Magnetics, kinetics and disruption alarm and fast shutdown systems are dedicated controllers consisting of several elementary control units (ECUs) devoted to the control of specific plasma parameters. The roles, present design status, and some remaining issues for the magnetics control system were described in Section 4. It is mentioned that continuous R&D works are required during the entire KSTAR project period for the embodiment of the PCS concept described in the present paper. A detailed design of software and hardware components of the KSTAR PCS, including the development of a prototype KSTAR PCS test module, is under progress and will be reported in the near future.
[4] [5]
[6]
[7]
[8]
[9] [10]
[11]
[12]
[13]
[14]
[15]
Acknowledgments [16]
The authors gratefully acknowledge fruitful discussions with Dr. J.B. Lister at CRPP-EPFL, Drs. G. Raupp and G. Neu at IPP-Garching and R. Chestnut at SLAC during their visits to KBSI for KSTAR control system review. This work was supported by the Korea Ministry of Science and Technology under the KSTAR project contract.
[17]
[18]
References [19] [1] G.S. Lee, J. Kim, S.M. Hwang, C.S. Chang, H.Y. Chang, M.H. Cho, et al., The KSTAR project: an advanced steady state superconducting tokamak experiment, Nucl. Fusion 40 (3Y) (2000) 575–582. [2] E.A. Lazarus, J.B. Lister, G.H. Neilson, Control of the vertical instability in tokamaks, Nucl. Fusion 30 (1990) 111–141. [3] J.B. Lister, E.A. Lazarus, A.G. Kellman, J.-M. Morret, J.R. Ferron, F.J. Helton, et al., Experimental study of the vertical stabil-
[20]
[21]
ity of high decay index plasmas in the DIII-D tokamak, Nucl. Fusion 30 (1990) 2349–2366. F. Hoffmann, S.C. Jardin, Plasma shape and position control in highly elongated tokamaks, Nucl. Fusion 30 (1990) 2013. F. Albanese, G. Ambrosino, E. Coccorese, F.C. Morabito, A. Pironti, G. Rubinacci, et al., Plasma current, shape and position control in ITER, Fusion Technol. 30 (1996) 167–183. J. Lister, A. Sharma, D.J.N. Limebeer, Y. Nakamura, J.P. Wainwright, R. Yoshino, Plasma equilibrium response modeling and validation on JT-60U, Nucl. Fusion 42 (2002) 708–724. F. Villone, P. Vyas, J. Lister, R. Albanese, Comparison of the CREATE-L plasma response model with TCV limited discharges, Nucl. Fusion 37 (1997) 1395–1410. M. Ariola, G. Ambrosino, J.B. Lister, A. Pironti, F. Villone, P. Vyas, et al., A modern plasma controller tested on the TCV tokamak, Fusion Technol. (1999) 126–138. http://www.aps.anl.gov/epics. T.W. Fredian, J.A. Stillerman, MDSplus: current developments and future directions, Fusion Eng. Des. 60 (2002) 229–233. P. Sichta, J. Dong, Startup of the EPICS at NSTX, in: Proceedings of the 18th Symposium on Fusion Engineering, Albuquerque, New Mexico, 1999. J.B. Lister, F. Hofmann, J.-M. Moret, F. Buhlmann, M.J. Dutch, et al., The control of tokamak configuration variable plasmas, Fusion Technol. 32 (1997) 321–373. K. Kurihara, T. Kimura, I. Yonekawa, J. Takahashi, M. Matsukawa, M. Mutoh, et al., JT-60 plasma control system, Fusion Eng. Des. 11 (1990) 441–454. T. Kimura, K. Kurihara, Y. Kawamata, K. Akiba, M. Takahashi, T. Terakado, et al., JT-60U plasma control system, Fusion Technol. 32 (1997) 404–415. M.L. Walker, D.A. Humphreys, J.A. Leuer, J.R. Ferron, B.G. Penaflor, Implementation of model-based multivariable control on DIII-D, Fusion Eng. Des. 56–57 (2001) 727–731. ITER Physics Expert Group on Disruptions, Plasma Control, and MHD, ITER Physics Expert Group on Energetic Particles, Heating and Current Drive, ITER Physics Expert Group on Diagnostics, ITER Physics Basis Editors, Plasma operation and control, Nucl. Fusion 39 (1999) 2578–2625 (Chapter 8). G. Raupp, K. Behler, R. Cole, K. Luddecke, V. Mertens, G. Neu, et al., ASDEX upgrade team: next generation discharge control system for ASDEX upgrade, Fusion Eng. Des. 46 (1999) 347–354. G. Raupp, K. Behler, G. Neu, R. Merkel, W. Treutterer, D. Zasche, et al., ASDEX upgrade team, integrating discharge control and data acquisition at ASDEX upgrade, Fusion Eng. Des. 60 (2002) 353–359. http://www.geindustrial.com/cwc/gefanuc/embedded/products/ hw rtnet pmc 5565.html. A. Chertovskikh, Hogun Jhang, M.C. Kyum, The hardware configuration for the plasma control system of the KSTAR tokamak, in Proceedings of the KSTAR Conference 2004, Muju, Korea, 2004. F. Hoffman, G. Tonetti, Fast identification of plasma boundary and X-points in elongated tokamaks, Nucl. Fusion 28 (1988) 519–522.
H. Jhang, I.S. Choi / Fusion Engineering and Design 73 (2005) 35–49 [22] W.H. Press, B.P. Flannery, S.A. Teukolsky, W.T. Vetterling, Numerical Recipes, Cambridge University Press, New York, 1986. [23] S.C. Jardin, Time scales for non-inductive current buildup in low aspect ratio toroidal geometry, Nucl. Fusion 40 (2000) 1101–1112. [24] S.G. Lee, J.G. Bak, private communication. [25] G.S. Lee, D.P. Ivanov, H.L. Yang, Hagun Jhang, J.Y. Kim, et al., Advanced physics and plasma control with segmented in-vessel control coils in the KSTAR tokamak, J. Plasma Fusion Res. 5 (2001) 261–264. [26] Hagun Jhang, C. Kessel, N. Pomphrey, S.C. Jardin, G.S. Lee, C.S. Chang, Design calculations for fast plasma position control in Korea Superconducting Tokamak Advanced Research, Fusion Eng. Des. 45 (1999) 101–115. [27] S.C. Jardin, N. Pomphrey, J. Delucia, Dynamic modeling of transport and positional control of tokamaks, J. Comput. Phys. 66 (1986) 481–507.
49
[28] Hogun Jhang, J.Y. Kim, Magnetics control simulations for KSTA plasmas, Fusion Eng. Des. 60 (2002) 397–401. [29] Hogun Jhang, C.E. Kessel, N. Pomphrey, J.Y. Kim, S.C. Jardin, G.S. Lee, Simulation studies of plasma shape identification and control in Korea Superconducting Tokamak Advanced Research, Fusion Eng. Des. 54 (2001) 117–134. [30] E.J. Strait, Stability of high beta tokamak plasmas, Phys. Plasmas 1 (1984) 1415–1431. [31] C.M. Bishop, An intelligent shell for the toroidal pinch, Plasma Phys. Control. Fusion 31 (1989) 1179–1189. [32] M.F.F. Nave, J.A. Wesson, Mode locking in tokamaks, Nucl. Fusion 36 (1996) 273–282. [33] R.J. La Haye, Physics of Locked Modes in ITER, Error Field Limits, Rotation for Obviation and Measurement of Field Errors, General Atomics Report GA-A22468, 1997. [34] A.H. Boozer, Resistive wall modes and error field amplification, Phys. Plasmas 10 (2003) 1458–1467.
