The design of the residual ion dump power supply for ITER neutral beam injector

The design of the residual ion dump power supply for ITER neutral beam injector

G Model ARTICLE IN PRESS FUSION-9442; No. of Pages 4 Fusion Engineering and Design xxx (2017) xxx–xxx Contents lists available at ScienceDirect F...

948KB Sizes 2 Downloads 25 Views

G Model


FUSION-9442; No. of Pages 4

Fusion Engineering and Design xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Fusion Engineering and Design journal homepage:

The design of the residual ion dump power supply for ITER neutral beam injector Alberto Ferro a,∗ , Elena Gaio a , Luca Sita b , Luigi Rinaldi b , Giovanni Corbucci b , Giuseppe Taddia b , Daniel Gutierrez c , Muriel Simon c , Hans Decamps d a

Consorzio RFX, Padua, Italy OCEM Energy Technology srl, Bologna, Italy c Fusion for Energy, Barcelona, Spain d ITER Organization, Saint Paul lez Durance, France b

h i g h l i g h t s • • • • •

RIDPS feeds the Residual Ion Dump of MITICA and ITER Heating Neutral Beam Injectors. RIDPS shall provide 25 kVdc + 5 kVac, low ripple, nominal current 60 A for 1 h. RIDPS is a Pulse-Step-Modulator (PSM) with features to enhance reliability and EMI. PSM guarantees high accuracy, redundancy and low ripple with small output filter. This implies low energy transferred to possible arcs between grid plates of RID.

a r t i c l e

i n f o

Article history: Received 3 October 2016 Received in revised form 24 March 2017 Accepted 22 April 2017 Available online xxx Keywords: Power supply Pulse step modulation Residual ion dump ITER NBI MITICA

a b s t r a c t The Residual Ion Dump Power Supply (RIDPS) is part of the Ground Related Power Supplies (GRPS) of the MITICA experiment of the ITER Neutral Beam Test Facility (NBTF) and the two ITER Heating Neutral Beam Injectors (HNBI). The GRPS will be manufactured by OCEM Energy Technology s.r.l. (OCEM) via a procurement contract with F4E. The RIDPS is devoted to feed the electrostatic Residual Ion Dump (RID), which deflects and collects the beam residual ions after the neutralization process. The maximum average voltage of the RIDPS is 25 kV, to which can be superimposed a sinusoidal or trapezoidal alternate voltage at 50 Hz, 5 kV maximum. The nominal current is 60 A, with a maximum pulse length of 1 h. This paper describes the detailed design of the RIDPS, highlighting its peculiar aspects, and the expected performance resulting from simulations. © 2017 Consorzio RFX. Published by Elsevier B.V.

1. Introduction The Ground Related Power Supplies (GRPS) for the two Heating Neutral Beam Injectors (HNBI) of ITER (Cadarache, France) will include the Residual Ion Dump Power Supply (RIDPS). The RIDPS is devoted to polarize the plates of the Residual Ion Dump (RID), the beam line component which has to deflect and collect the residual ions downstream the Neutralizer [1]. The same power supply will be also provided for MITICA (Megavolt ITER Injector & Concept Advancement), the full-scale prototype of the ITER HNBI, under construction in Padua (Italy) [2].

∗ Corresponding author. E-mail address: [email protected] (A. Ferro).

The RIDPS has to provide a voltage which is the sum of a dc component (ranging between 5 and 25 kV) plus an ac component (maximum 5 kV peak), thus the maximum peak output voltage is 30 kV [3]. The voltage ripple has to be lower than ±500 V and the maximum current is 60 A. In addition, the RIDPS has to include a fast protection against short-circuits at the output, due to possible arcs which could occur between RID plates; in such events, the arc energy has to be limited as much as possible, to avoid damages. Different design solutions have been studied in the past for the RIDPS. For example, in [3] it was proposed a topology based on the series connection of a thyristor rectifier and a multilevel stage with 8 H-bridges. Despite some advantages, this topology was discarded because, due to the stringent ripple requirement, an important output filter would have been necessary; in case of arcs, its stored energy would be discharged on the RID plates. Therefore, a new 0920-3796/© 2017 Consorzio RFX. Published by Elsevier B.V.

Please cite this article in press as: A. Ferro, et al., The design of the residual ion dump power supply for ITER neutral beam injector, Fusion Eng. Des. (2017),

G Model FUSION-9442; No. of Pages 4



A. Ferro et al. / Fusion Engineering and Design xxx (2017) xxx–xxx

Table 1 Main requirements of the RIDPS. Input voltage Maximum value of dc output voltage component Max. peak value of ac output voltage component Maximum output current Ac voltage waveforms Period of ac voltage Nominal rise time of trapezoidal ac voltage Voltage regulation range Maximum voltage ripple Maximum voltage fluctuation under transients Voltage regulation accuracy in steady-state Max. rise time of the output voltage from 0 to 30 kV Maximum pulse length (t) Duty cycle Pulses during lifetime

22 kVrms ± 10%, 50 Hz 25 kV 5 kV 60 A Trapezoidal/Sinusoidal 20 ms 2 ms (peak to peak) 0 ÷ 100% ±500 V ±2500 V (with 60 A step variation of output current) ±1250 V (ripple excluded) 200 ms (with any load current, up to 60 A) 1h t < 150 s: 1 shot every 10 t > 150 s: 25% 50000

reference design has been developed, based on the Pulse Step Modulation (PSM) technology [4]. The advantages are: high modularity, high accuracy and, most of all, the absence of a large capacitance at the output. With PSM, in fact, thanks to the high number of modules in series and the phase-shift of the respective Pulse Width Modulation (PWM) carriers, the ripple is very limited and the output voltage regulation is very fast, thus the output filter can be much reduced. The main disadvantage of the PSM solution is the need of a transformer with many secondary windings. However, the present technology of cast resin transformers allows manufacturing such device without major issues. The PSM technology is widely used in fusion applications, however the RIDPS presents specific features which allow reducing the electromagnetic emissions and improving its reliability and availability. The selected Industrial Supplier endorsed the PSM solution; the detailed design is being developed starting from the same technology adopted for other two power supplies provided by the same company for SPIDER ISEPS (EGPS and RFPS, [5]). After a brief overview of the technical requirements, this paper describes the present design of the RIDPS. 2. Main requirements The main requirements of the RIDPS are reported in Table 1. The load is represented by the RID, where the residual beam ions are deflected by the electrostatic field generated by five panels and damped on them. The field is produced by applying positive or negative potential to the two even panels, while the others are grounded locally. The RIDPS output current is essentially imposed by the beam source, with little dependence on the output voltage. Therefore, the load of the RIDPS can be approximated as a current generator. A changeover facility at RIDPS output is required to invert the voltage at RID plates, in order to test both polarities in MITICA. Therefore, both RIDPS output terminals have to be insulated for 30 kV with respect to ground. In MITICA, the RIDPS will be connected to the 22 kV ac distribution grid by the respective withdrawable circuit breaker foreseen in the MITICA Distribution Board. At ITER Site, instead, the RIDPS shall also include a 24 kV ac disconnector and earthing switch, to be connected to the ITER 22 kV ac distribution grid. In MITICA, the RIDPS will be installed in the MITICA Power Supply hall (Building 3), composed of two floors. The power section will be installed at the ground floor, while the Local Control System will be installed at the first floor. At ITER Site, the RIDPS of both HNBI1 and HNBI2 will be installed in Building 34, organized in two

Fig. 1. RIDPS mechanical layout.

floors as in MITICA. Both Building 3 and Building 34 are far from the Neutral Beam Injector, therefore the output cable of RIDPS will be very long (in MITICA the cable length is about 150 m). This unusual situation requires the adoption of a coaxial output cable and the common-mode output filter described in the following. 3. Power section design The topology adopted for the RIDPS is the PSM [4], with 42 power modules in series at the output, supplied by a multi-secondary cast resin transformer. Half secondary windings are star and half are delta connected, giving a current absorption from the mains with 12 pulses. The power modules are arranged in six columns and seven rows in a plastic frame, as shown in Fig. 1. Both transformer and modules are mounted on a common iron pedestal, allowing the transportation of the whole RIDPS without disassembling the 126 power cables connecting transformer and modules (Fig. 1). Therefore, the insulation test of the power modules will be performed in factory only, at 60 kVdc with an high voltage generator. All the power modules are cooled with demineralized water by means of a couple of distribution pipes, respectively at the bottom (inlet) and at the top (return) of the frame, feeding the six module columns in parallel. The water flow on each column is monitored through a flow-switch. Thanks to the high module efficiency, the total required flow rate is less than 0.5 L/s at 10 bars with a maximum water temperature increase of 8 ◦ C. The overall electric scheme of the RIDPS is shown in Fig. 2. The IGBTs of the power modules are commutated with a proprietary modulation algorithm called Multi Pulse Width Modulation (MPWM) [6]. The ripple frequency at the output is fixed at 84 kHz. Thanks to the phase-shift of the PWM carriers of the modules, their frequency is 2 kHz only (with 42 working modules), which is suitable for the IGBTs. The voltage ripple at the output is smoothed by a low-pass RLC output filter with the values of Fig. 2, where the inductance is a custom-made air-wound inductor and the capacitance is made of 25 capacitors (each rated for 1250 V) in series. At the return point of the output cable, at RIDPS side, an additional RC filter to ground is foreseen to shunt the common mode currents,

Please cite this article in press as: A. Ferro, et al., The design of the residual ion dump power supply for ITER neutral beam injector, Fusion Eng. Des. (2017),

G Model FUSION-9442; No. of Pages 4

ARTICLE IN PRESS A. Ferro et al. / Fusion Engineering and Design xxx (2017) xxx–xxx


Fig. 2. Scheme of the RIDPS.

Fig. 3. Simplified scheme of a power module.

avoiding them being transferred toward the grounding connection at load side, causing electromagnetic disturbances. To achieve the high reliability required for this application, two redundant measurement chains have been adopted for output voltage and current, one for closed control loop and the other for protection purpose. 3.1. Power transformer The 22 kV/42 × 633 V transformer is the most critical device of the RIDPS, due to the high number of secondary windings and the high dc voltage of the secondary terminals versus ground (up to 30 kV). The nominal power of the transformer is 1.9 MVA. To absorb from the main a current equivalent to that of a 12 pulse rectifier, half of the secondary windings are delta connected and half are star connected. A fuse in series to each secondary winding is foreseen; its aim is to protect the transformer against overloads that, if occur only in a few secondary windings, cannot be detected from the protections at the primary side. A screen is interposed between primary and secondary windings, connected to ground. The estimated dimensions of the transformer are about 2.5 × 1.5 × 3.5 m (W × D × H).

power transformer is energized, Q2a is closed, Q2b (complementary to Q2a) is open and the dc-link is charged through the inrush current limiting resistor R1 = 39 . When the proper dc-link voltage level is reached, Q1 is closed, shorting R1. This solution allows charging the capacitors directly at the transformer energization, without a separate pre-charge network. The dc-link filter has the following functions: (1) to operate the rectifier in Continuous Conduction Mode (CCM), so as to improve the power factor [7]; (2) to reduce the 300 Hz rectification ripple of the voltage; (3) to filter the ac current ripple due to the IGBT switching; (4) to reduce the dc-link voltage oscillations at load transients. The capacitance of the filter (C1 + C2) is 3100 ␮F, storing 1.12 kJ at 850 V. Due to this high energy, for safety reasons, besides the permanent discharging resistance R4, an additional branch for automatic fast discharge has been included (Q2b and R3). In addition, a LED warns about the presence of voltage on the capacitors. If the module is faulty or switched off, the switch Q3 shortcircuits its output terminals. This guarantees the continuity of service in case of fault without removing/shunting the module. To avoid damages, Q3 is closed by the control board only if the output voltage is sufficiently low. Each power module includes a water-cooled heat-sink, which accommodates the diode bridge, the IGBT module, R3 and R4. To remain within the maximum junction temperature with a certain margin and to keep the water temperature at the outlet below 43 ◦ C, its thermal resistance will be lower than 0.16 K/W. The maximum total power loss of the RIDPS dissipated in water is 15 kW. The layout of the power module is very compact and has been derived from that one developed for SPIDER-EGPS [5]. The ac input and the dc output connections are realized with Multi-Contact connectors, while, for the water, fast no-spill connectors are adopted.

4. Control section design 3.2. Power modules The number of modules (42) has been chosen considering the maximum output voltage to be produced (30 kV), a suitable nominal dc-link voltage (850 V), the minimum ac input voltage, the voltage drops in the power switches and keeping a margin of 2 modules to improve system availability in case of faults. As shown in Fig. 3, each power module will be basically composed of a 3-phase diode bridge (based on three modules Semikron SKKD 81/22 H4), a LC damped low-pass filter as dc-link and a phase-leg IGBT module (Semikron SKM200GB17E4) acting as a buck dc-dc converter. The nominal dc-link voltage has been selected in order to use the widely available and reliable IGBT series having collector-emitter blocking voltage (VCES ) of 1700 V, keeping the required high voltage margin. When the power module is off, Q2b (which is a NC switch) is closed and the dc-link is discharged through R3 = 700 . When the

The control section is composed of a “Fast Control System” and a “Slow Control System”. The first is based on a set of custom Field Programmable Gate Arrays (FPGA) boards, with these main functions: (1) to acquire the measurements of the output voltage and current from the 10 MS/s A/D converter boards with 2 GBd optical links; (2) to acquire the measurements of the step-down transformer primary voltages and currents from the 200 kS/s A/D converter boards; (3) to implement the PI voltage regulator and the modulation algorithms for the power modules; (4) to communicate with the power modules via 10 MBd optical links; (5) to communicate reference and measurements to the HNBI Plant System Control; (6) to realize the fast protections, in particular, against overcurrents and overvoltages at the output. The Slow Control System is based on a PLC Siemens S7 400 and its main functions are to manage the slow control logic and alarms and to implement

Please cite this article in press as: A. Ferro, et al., The design of the residual ion dump power supply for ITER neutral beam injector, Fusion Eng. Des. (2017),

G Model FUSION-9442; No. of Pages 4



A. Ferro et al. / Fusion Engineering and Design xxx (2017) xxx–xxx

components, allowing the choice of the proper devices. The output voltage ripple has been quantified too, resulting less than ±280 V in nominal conditions, much lower than required. Load variations have been simulated; an example is shown in Fig. 5. Here, a load current variation from 0 to 60 A with slope 800 A/ms is assumed. As it can be noted, the corresponding voltage undershoot is almost negligible. Finally, an arc event and consequent trip of the overcurrent protection have been simulated, assuming a current threshold set to 85 A and a latency between the protection trip and the actual IGBT switch-off equal to 2.5 ␮s. The resulting energy into the arc is 37 J (assuming an arc voltage of 100 V), with a current peak of 590 A: low values, thanks to the selected technology. ®

Fig. 4. PLECS model of the RIDPS, main schematic.

6. Conclusions The design of the RIDPS is well advanced. The scheme and the main parameters have been already chosen and the main components have been selected. Simulations with detailed models have been run and the results confirm the excellent performance achievable with the PSM topology and the technology available today. In particular, thanks to the high number of modules (42), the equivalent switching frequency at the output can be set to a very high value (84 kHz), giving a minimal voltage ripple without using a large output filter. This in turn implies a low charge and energy transferred to a possible arc between the grid plates of the RID, reducing the possibility of damages. Particular provisions have been taken to increase the reliability of the system, e.g. two redundant chains for output measurements, the possibility to operate with two faulty power modules without any manual intervention and the selection of power switches with high safety margins.

Fig. 5. Output voltage and current derived from simulation.

the state machine, synchronized with the HNBI Plant System Control via Ethernet. Fast and Slow Control Systems communicate via RS232. The control section is physically divided into a Local Control System (LCS) cubicle at first floor (with the Human-Machine Interface based on a touch screen panel), a peripheral part close to Medium Voltage measurements and a peripheral part close to power modules. Each power module is equipped with a dedicated control board with microprocessor, which supervises and commands the module devices, acquires the measurements of dc-link, auxiliary and output voltages and manages the internal faults. Each control card communicates with the LCS through a couple of optic fibers: one drives directly the IGBTs, the other gives back the module status, alarms and measurements. 5. Simulations ®

The RIDPS has been modelled and simulated using PLECS , able to integrate control and power circuits. An overall view of the model is shown in Fig. 4. The “Regulator” block contains a discrete integral regulator with overcurrent protection, the “Modulator” includes 42 phase-shifted MPWM modulators and the “Modules” contains 42 modules having the scheme of Fig. 3. The “Outer Circuit” includes the output filter and an ideal current generator representing the load. This model has been used to derive the current flowing in the

Acknowledgment The project F4E-OPE-278 Lot 2 has been funded with support from Fusion for Energy. This publication reflects the views only of the author, and F4E cannot be held responsible for any use which may be made of the information contained therein. The views and opinions expressed herein do not necessarily reflect those of the ITER Organization. References [1] ITER Technical Basis 2002 Detailed Design Document, section 5.3 (DDD5.3), IAEA, Vienna. [2] V. Toigo, et al., Progress in the realization of the PRIMA neutral beam test facility, Nucl. Fusion 55 (2015). [3] M. Barp, E. Gaio, V. Toigo, Studies on the ITER NBI residual ion dump power supply system, IEEE Trans. Plasma Sci. 40 (2012) 659–664. [4] N. Tomljenovic, W. Schminke, H.G. Mathews, Solid state dc power supplies for gyrotron and NBI sources, Proc. of 17th Symp. on Fus. Tech., Rome, Sept. (1992). [5] M. Bigi, et al., Design, manufacture and factory testing of the Ion Source and Extraction Power Supplies for the SPIDER experiment, Fus. Eng. and Des. 96–97 (2015) 405–410. [6] L. Rinaldi, A method for voltage modulation, International Patent WO2006/111840, Oct. 26th, 2006. [7] M.M. Jovanovic, D.E. Crow, Merits and limitations of full-bridge rectifier with LC filter in meeting IEC 1000-3-2 harmonic limit specifications, IEEE Trans. Ind. Appl. 33 (1997) 551–557.

Please cite this article in press as: A. Ferro, et al., The design of the residual ion dump power supply for ITER neutral beam injector, Fusion Eng. Des. (2017),