Fusion Engineering and Design 84 (2009) 2037–2041
Contents lists available at ScienceDirect
Fusion Engineering and Design journal homepage: www.elsevier.com/locate/fusengdes
Assessment of performance of the acceleration grid power supply of the ITER neutral beam injector Loris Zanotto ∗ , Alberto Ferro, Vanni Toigo Consorzio RFX, EURATOM-ENEA Association, Corso Stati Uniti 4, I35127 Padova, Italy
a r t i c l e
i n f o
Article history: Available online 30 December 2008 Keywords: ITER Neutral beam injector Power supply Inverter
a b s t r a c t This paper deals with the analyses carried out to assess the performance of the acceleration grid power supply of the ITER neutral beam injector; a model of the power supply is setup and simulations are run in both normal and anomalous operating conditions, with the aim of demonstrating the fulﬁlment of the requirements and to give inputs for the design of the equipments. The dynamic performance and the output voltage ripple are studied in normal operation on the basis of a conceptual design of the system; anomalous operating conditions, such as breakdown of the grids and sudden interruption of the beam current, are also analyzed and discussed. © 2008 Elsevier B.V. All rights reserved.
1. Introduction The ITER neutral beam injector (NBI) is designed to deliver up to 16.5 MW of additional heating to the plasma, accelerating deuterium and hydrogen negative ions up to −1 MV with a beam current as high as 40 A (deuterium operation) and 46 A (hydrogen operation), for as long as 3600 s . The beam is created and accelerated by means of two main power supplies: the acceleration grid power supply (AGPS), providing power to the acceleration grids, and the ion source and extraction grid power supplies (ISEPS), devoted to supply the ion source components and the extraction grid. The ion production is based on a radio-frequency-driven ion source and the 1 MeV accelerator, named MAMuG (multi aperture multi grid), is composed of ﬁve acceleration gaps, of −200 kV each . The design of acceleration grid power supplies evolved from early NB systems based on thyristor converters, with comparatively slow response, acceleration voltages up to 160 kV dc and tetrode switches placed in series to protect the grids against breakdown. Now, new schemes are adopted, based on modular conﬁgurations that comprise solid state inverters placed upstream of insulating step-up transformers feeding the grids ; in these schemes the protection against breakdown relies on fast switch-off of the solid state inverters. Examples of such power supplies are found in the −500 kV AGPS of the JT-60U negative ion NBI  and in the recent upgrade of the JET PINI system . With respect to JT-60U, the ITER NBI AGPS, which has a similar scheme, is requested to deliver more power (about 50 MW) to the acceleration grids with a remarkably increased acceleration
∗ Corresponding author. Tel.: +39 049 829 5898; fax: +39 049 870 0718. E-mail address: [email protected]
(L. Zanotto). 0920-3796/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.fusengdes.2008.11.020
voltage (−1 MV). The design is complicated by the expected regular occurrence of acceleration grid breakdowns, corresponding to quick and repetitive application of load short circuits during full power operation. The breakdowns constitute, rather than exceptional faults, routine events that the system must be capable of handling without damage to the grids and with no deterioration of the voltage holding capability. These requirements, together with other speciﬁcations concerning the dynamic response and precision, addressed the design of the ITER NBI AGPS towards a solution which is near the limit of modern available technologies in the area of high power static conversion systems and of the very high dc voltage equipments. The complexity of the project is also related to the framework of the ITER implementing agreement, which foresees a sharing between European Union (EU) and Japan (JA) domestic agencies on the procurement of NBI components and subsystems. As a consequence, the AGPS will be partially provided by EU, which will deliver the low voltage parts, and by JA, providing the high voltage components. An interface within the AGPS is therefore established, and the deﬁnition of a set of agreed interface parameters becomes crucial for an integrated design of the system. For the mentioned reasons, the availability of a tool for simulating, assessing and analyzing the performance of the AGPS is essential to assist the design and verify the compliance with the requirements, both in normal and abnormal conditions and with different selection of interface parameters values. To this end, a detailed model of the AGPS power conversion system was built and the operation simulated in various scenarios. This model was used to select the interface parameters, to perform an integrated conceptual design of the AGPS and to assess an optimized control strategy. After giving an overview of the AGPS basic scheme, operation and requirements, this paper will focus on the implemented model
L. Zanotto et al. / Fusion Engineering and Design 84 (2009) 2037–2041
Fig. 1. Scheme of the AGPS. Current and voltage values refer to deuterium operation.
and discuss the results of the simulations carried out to address some peculiar aspects of the AGPS operation. In particular, the dynamic performance of the system will be analyzed and the output voltage ripple evaluated as a function of some interface parameters. Moreover, the case of grids breakdown will be investigated from the point of view of the conversion system; ﬁnally, the behaviour of the AGPS at the sudden interruption of the beam current (beam off), will be analyzed to determine the maximum voltage stress on the system. 2. The ITER NBI acceleration grid power supply scheme and requirements 2.1. Scheme and basic requirements The reference scheme of the AGPS is shown in Fig. 1. It consists of a step-down transformer feeding an ac/dc conversion system, which at the output side is connected to ﬁve neutral point clamped (NPC), three phase inverters  based on solid state switches. A principle scheme of a NPC three phase inverter is represented in Fig. 2. The inverters are fed by a common dc-link and supply a step-up transformer each. The step-up transformer feeds a diode rectiﬁer; the rectiﬁers are connected in series at the output side to obtain the nominal acceleration voltage of −1 MV for deuterium operation (−870 kV for hydrogen). Since negative ions shall be extracted and accelerated, the acceleration voltage must be negative; therefore, the power supply positive polarity is connected to the beam grounding and the negative polarity is linked up with
the plasma source. Of course, the high voltage side of the AGPS (DC generator, DCG) must be insulated from the low voltage side, commonly identiﬁed as the AGPS Conversion System; this is accomplished with the step-up transformers, which must be designed to withstand dc voltages to ground decreasing from −200 kV (DCG5 stage) to −1 MV (DCG1 stage). The AGPS scheme is completed by ﬁve dc ﬁlters, placed immediately downstream of the diode rectiﬁers, mainly to provide a limitation of the output voltage ripple and of the voltage at beam off. Diode rectiﬁers and ﬁlters are placed inside tanks and insulated with SF6 gas. The power supply is connected to the acceleration grids by means of a −1 MV dc transmission line insulated with SF6 gas, comprising intermediate potential conductors to supply the intermediate grids, about 100 m long. The AGPS load is represented by the beam current, created and controlled by the ISEPS; ISEPS and AGPS will control the beam current and the grid voltages so as Eq. (1) is satisﬁed, in order to assure the focusing of the beam (the so-called perveance matching conditions): k 1.5 Ibeam = √ Vacc m
where Vacc is the acceleration grid voltage and Ibeam is the beam current, k depends on the accelerator geometry and m on the accelerated species . The regulation of the output voltage is achieved by means of the inverter modulation. The inverters apply a square wave voltage at a ﬁxed frequency to the step-up transformers, with the possibility to change the duty cycle of the waveform, thus controlling the rectiﬁed voltage amplitude at the output. An example of output phase voltage of a NPC inverter for two different modulation indexes is represented in Fig. 3. The grid currents and voltages required from the AGPS are reported in Table 1, both for deuterium and hydrogen operation. The total rated power delivered by the AGPS is about 52 MW for both cases. Table 1 AGPS main output ratings (deuterium and hydrogen operation). Parameter
Main supply Grid 1 Grid 2 Grid 3 Grid 4 Grounded conductora
Fig. 2. Principle scheme of a NPC three phase inverter.
D− , 1 MeV, 40 A
H− , 870 keV, 46 A
−1000 −800 −600 −400 −200 0
59.4 3.0 5.3 3.1 2.7 45.3
−870 −696 −522 −348 −174 0
62.8 2.4 4.3 3.1 2.1 50.9
a Beam current + accelerated electrons at the output of the grounded grid + grounded grid current.
L. Zanotto et al. / Fusion Engineering and Design 84 (2009) 2037–2041
Fig. 3. Example of output phase voltage of a NPC inverter. Two cases with different modulation indexes (m) are represented.
2.2. Interface parameters
capacitance at breakdown; its value is not high enough to damp the ﬁlter.
The interface between EU and JA procurements in the AGPS scheme lies between the AGPS Conversion System (to be procured by EU) and the DC generator (JA). Since the procurements are strongly integrated, a large effort was spent to identify a set of electrical interface parameters to be jointly deﬁned and agreed in order to mutually satisfy the feasibility constraints of both procurements and, at the same time, optimize the design of the overall AGPS. The most important interface parameters are the inverter switching frequency, the dc-link voltage, the overall inductance seen by the inverters (which comprises the per-unit short-circuit inductance of the step-up transformers and any additional inductance placed at the output of the inverters) and the shape of the inverters output voltage. All interface parameters affect more than one aspect of the design, and the choice of each parameter can be performed only after a comprehensive and integrated study of the system ; here we focus on the inverter switching frequency and on the dc ﬁlter parameters, which have a strong inﬂuence on the dynamic performance and on the output voltage ripple. The best frequency value derives from a trade-off between different constraints. Loss reduction in the active switches of the inverters, feasibility of the step-up transformers and limitation of the overcurrent at breakdown ask for a frequency as low as possible. On the other hand, good dynamic performance, reduction of the over-voltage at beam off and output voltage ripple limitation require a switching frequency as high as possible. The assessment of the switching frequency value was done in  with the main criteria of limiting the output voltage ripple and considering the feasibility of inverters and transformers. Two values were selected for further analyses to be done with a model of the AGPS: 150 and 300 Hz. A further contribution to the ripple reduction comes from the dc ﬁlters; the ﬁltering effect is due to the combination of the ﬁlter capacitance and of the total inductance upstream of the ﬁlter (transformer short-circuit inductance plus additional inductance in series to the inverter outputs). However the capacitance has been introduced with the main aim of limiting the system over-voltage at load beam-off, due to the energy stored in the upstream inductances and to the delay in switching off the inverters; at the same time, the capacitance value must be limited to avoid the discharge of excessive energy on the acceleration grids at breakdown. A rough estimation of the over-voltage at beam off is given by Eq. (2): Vout ∼ =
Inom tdelay C
where Vout is the over-voltage of one stage, Inom the nominal dc current of the stage, C the ﬁlter capacitance and tdelay is the time interval needed to switch-off the inverters. Assuming C = 300 nF and a reasonable cut-off time of the inverters (150 s), the estimation gives an over-voltage below 20%. The dc ﬁlters are also equipped with a resistance connected in series to the capacitance to limit the peak current due to the discharge of the energy stored in the
2.3. Dynamic and precision requirements The AGPS is required to operate within a voltage range from 20% to 100% of the nominal value, guaranteeing the correct voltage sharing between grids and the voltage stability value for long operational time (on the order of 1 h). A relatively slow modulation of the voltage may be required within the mentioned range, to regulate the beam energy as required in the various phases of the ITER discharge. A particular set of requirements concerns the precision and the dynamic at the beam turn-on after a breakdown of the grids. Such requirements are summarized in Table 2. 2.4. Anomalous operating conditions On top of the requirements pertaining to the normal operation, the AGPS must be able to handle the anomalous operating conditions, i.e. the breakdown of the grids and the beam off. In the ﬁrst case, the AGPS conversion system has to limit the delivery of extra energy to the grids in excess of the energy stored in the downstream capacitances; an obvious additional requisite is that the breakdown must not damage the power supply. The protection is performed at the low voltage side with fast switching off of the inverters. The energy stored in the ﬁlter and in the stray capacitances cannot be recovered back to the dc-link; therefore passive elements must conveniently be placed in the transmission line, e.g. resistors and core snubbers to partially dissipate this energy . In case of breakdown, the AGPS control must switch off the inverters within 150 s from the breakdown (cut-off time) and, after a suitable time (20 ms), the voltage must be reapplied with an appropriate ramp up to the set-point, with the dynamic and precision already described in the previous section. Concerning the beam off event, occurring when the beam current expires abruptly without contemporary AGPS shutdown, the AGPS control must be able to quickly switch off in order to limit the arising over-voltage at the output, which could be very dangerous for the integrity of the high voltage components. Table 2 AGPS dynamic and precision requirements. Parameter
D− or H− operation
Voltage regulation range Voltage resolution Output voltage accuracy for 1 h operation Maximum voltage ﬂuctuation Maximum voltage ripplea Rise time of the output voltageb Maximum settling timec
20–100% 1 kV ±2% ±2.5% at ﬂat top ±5% 80 ms 50 ms
a b c
It is referred to the average output voltage in steady state condition. Time to reach 90% of the full voltage. Time from 90% of the full voltage within the accuracy range.
L. Zanotto et al. / Fusion Engineering and Design 84 (2009) 2037–2041
3. The AGPS model
4. Result of the simulations
The AGPS model was setup with PSIMTM software . Depending on the purpose of the analyses, the models were customized with different levels of detail. The analyses of the dynamic performance and of the ripple required to build-up a comprehensive model of the AGPS Conversion System, including a rather detailed reproduction of the step-down transformers, the ac/dc converter and the dc-link capacitor banks. Some assumptions were made on the structure of the power supply. The ac/dc converter was modeled as a 12-pulse thyristor converter made by two basic bridges connected in series at the output and supplied by a three windings transformer star/star/delta connected, to obtain the necessary 30◦ of phase shift. The dc-link was arranged with a balanced central point needed for the connection of the neutral point of the inverters. The dc-link voltage value was chosen as a compromise between different feasibility issues; a high dc-link voltage was suggested to reduce the output current of the ac/dc converter and to increase the step-up transformer turn ratio, both deemed critical for the design; conversely, the feasibility of the inverters required that the dc-link voltage be not excessive. In the model, a design based on integrated gate commutated thyristor (IGCT) components  has been arbitrarily assumed; a reasonable total dc-link voltage value for this kind of components now available on the market is 6.5 kV. The dc-link capacitance is distributed among the stages and its equivalent value was set to 10 mF per stage at the full dc-link voltage, giving a global dc-link capacitance of 50 mF. This number was estimated on the basis of a rough calculation aimed at limiting the dc-link voltage drop at inverter on and the dc-link over-voltage at inverter off within 10%. The stepup transformers were implemented assuming design parameters as reported in ; the turn-ratio is calculated to ensure the output nominal voltage (−1 MV) in all possible conditions, including a reduction of the ac network grid voltage. The choice of the turnratio is linked also to the per-unit short-circuit voltage (vsc%) of the transformers. A value of vsc% = 15% was ﬁnally agreed between EU and JA, and assumed in the model. The transmission line, connected downstream of the diode rectiﬁers and of the ﬁlters, was modeled using a simpliﬁed lumped parameter approach, accounting for all capacitive couplings between conductors and for the passive protections (core snubbers and series resistors). The use of a more sophisticated model of the line  was beyond the scope of the analysis and too computation time consuming. The load of the AGPS was represented by ideal current generators connected between the grids and the grounded grid. The current values during the startup of the beam are controlled as a function of the acceleration voltage to maintain perveance matching conditions, as described in Eq. (1). As far as the control is concerned, the inverters are feedback controlled on the output voltage, with a feed-forward component based on the output voltage reference waveform. The ac/dc converter is feedback controlled on the dc-link voltage; the control loop is based on a proportional-integral regulator tuned to maximize the dynamic performance while maintaining the system within the stability margin. In order to save computational time, the full model described above was simpliﬁed to study the breakdown and beam off impact on the AGPS. Since the dynamic of the dc-link voltage was not important for the analysis of those events, the circuitry upstream of the inverters was reduced to two ideal voltage generators. The remaining part of the model has been left as in the full version. The simulation of the breakdown was performed by short-circuiting the load current generators with ideal switches; the arc among the grids is simply modeled with a 100 V ideal voltage generator. The beam off is simply simulated by turning-off the load current generators.
4.1. Dynamic performance and ripple The dynamic performance of the AGPS conversion system has been assessed running a simulation of the full model and comparing the results for the selected modulation frequencies (150 and 300 Hz); the simulations are useful to evaluate the output voltage ripple, also. The results for 150 Hz are shown in Fig. 4, where the total acceleration grid voltage, the voltage reference, the current streaming out from the grounded grid and the dc-link voltage are reported for deuterium operation at full power and in perveance matching conditions (results for hydrogen operation are similar). A 80 ms ramping acceleration voltage reference is required to the power supply, starting at 150 ms. At 350 ms, a complete breakdown on the ﬁve grids is applied, and all the inverter switches are turned off. The voltage is then reapplied after 20 ms as speciﬁed. It can be noted that the speciﬁcation on the rise time is not fully met at 150 Hz, as the voltage reaches 167 kV (84%) instead of the required 180 kV (90%); on the contrary, at 300 Hz the system complies with the rise time requirement. Anyway, the dynamic performance at 150 Hz can be improved by simply increasing the slope of the voltage request. The settling time requirement is achieved within the voltage accuracy range with both modulation frequencies; as expected the overall dynamic performance is better at 300 Hz. Concerning the output voltage ripple, the ﬂuctuations stay within ±5% of the rated voltage as required, 300 Hz being the frequency at which the ripple is lower, as expected. Note that the output voltage ripple does not comply the requirements at lower output voltages if the inverter modulation index is reduced; instead, as speciﬁc simulations not reported here have demonstrated , it is more convenient from this point of view to operate always at the maximum inverter modulation index. Therefore, a control technique to minimize the ripple can be implemented that regulates opportunely the dc-link voltage as a function of the output voltage reference, in order to make the inverters always working at the full modulation index. In this way, the output voltage ripple limitation was guaranteed at reduced output voltage also. The simulation results validated both values of the frequency and the selection of the dc ﬁlter parameters with respect to the ripple and the dynamic requirements. 150 Hz was ﬁnally chosen as it is less demanding for the design of the step-up transformers and of the inverters.
Fig. 4. Dynamic performance of the AGPS with an inverter switching frequency of 150 Hz. (a) Total acceleration grid voltage (continuous line) and reference voltage (dotted line). (b) Current streaming out from the grounded grid. (c) Inverter dc-link voltage.
L. Zanotto et al. / Fusion Engineering and Design 84 (2009) 2037–2041
short-circuit inductance of the transformers and from the inverters output inductances. Simulations were run with the aim of quantifying such over-voltage assuming that a protection is triggered conveniently, switching off the inverters. A delay of 150 s was assumed between the beam-off event and the inverter switch-off. The simulation results show that the peak voltages are different for the different stages and the total over-voltage is below 150 kV for deuterium operation. This is in line with the simpliﬁed calculation based on Eq. (2). The results conﬁrm the choice of the ﬁlter capacitance value; the output of the simulation was used to set the maximum voltage stress to be considered to design the AGPS high voltage components for this kind of anomalous condition. 5. Conclusions
Fig. 5. Simulation of a breakdown, with output current 66A (5% over the main supply current for the hydrogen operation) and trigger of the protection at the commutation of the inverter phase R. The phase voltages (a), the current in the external and internal upper switches of the inverter (b) and the output current (c) are reported.
4.2. Breakdown In case of breakdown, the AGPS conversion system must be switched off as quickly as possible to limit the energy delivered to the grids. The requirements to the protection system to guarantee the protection of the grids is that the cut-off time, deﬁned as the necessary time interval between the occurrence of a breakdown and the switch off of all 12 inverters switches of a stage, shall be less than 150 s. The cut-off time is the result of three main contributions: the detection time, spanning from the breakdown event to the protection request, estimated as 50 s, the switching time, needed for the commutation of a switch, and the minimum on time of the switch, deﬁned as the minimum time to be allowed for conduction before issuing a switch-off command. With IGCT components and on the basis of industrial experience last two contributions were estimated as 25 and 50 s respectively. Taking into account that the external switches of a NPC inverter leg must be switched off before the internal ones (otherwise the full dc-link voltage plus the transient commutation over-voltage may appear on the internal switches), the simulation revealed that the most critical condition as far as the cut-off time is concerned corresponds to a simultaneous commutation of two inverter legs. An example is reported in Fig. 5. Here, the breakdown protection is triggered just after the commutation of the external switch of the phase R. After 75 s from the breakdown (detection time + switching time) the phases S and T are switched off and the output current stops increasing. After 125 s from the breakdown (detection time + minimum on-time + switching time) the external switch of the phase R is turned off. Finally, after another switching time also the internal switch of the phase R is turned off, resulting in a total cut-off time of 150 s, matching the speciﬁcations. 4.3. Beam-off The consequences of an abrupt interruption of the beam current have been investigated. Qualitatively, the main expected effect is an over-voltage at the output due to the release of energy from the
The paper reported and discussed the analyses carried out to assess the performance of the ITER NBI AGPS. The development of a detailed simulation model of the system has proven to be a necessary step to evaluate the inﬂuence of the design parameters. An integrated conceptual design has been validated thanks to the model, analyzing the dynamic performance of the system and the output voltage ripple value, both resulting within the given requirements. The model has been used to study anomalous operating conditions also, i.e. breakdown of the acceleration grids and beam-off, further validating the choice of the design parameters. Acknowledgments This work, supported by the European Communities under the contract of Association between EURATOM/ENEA, was carried out within the framework the European Fusion Development Agreement. The views and opinions expressed herein do not necessarily reﬂect those of the European Commission. References  ITER technical basis, IAEA, Vienna 2002, ITER EDA, Doc. Series N. 24, Plant Description Document, Sec. 2.5.1.  R.S. Hemsworth, A. Tanga, V. Antoni, Status of the ITER neutral beam injection system (invited), Rev. Scientif. Inst., 79, February 2008.  E. Gaio, V. Toigo, A. De Lorenzi, R. Piovan, L. Zanotto, The Alternative design concept for the ion source power supply of the ITER neutral beam injector, Fusion Eng. Des. 83 (2008) 21–29.  M. Kuriyama, N. Akino, M. Araki, N. Ebisawa, M. Hanada, T. Inoue, et al., High energy negative-ion based neutral beam injector for JT-60U, Fusion Eng. Des. 26 (1995) 445–453.  D. Ganuza, J.M. Del Rio, I. Garcia, F. Garcia, P. Garcia de Madinabeitia, A. Perez, J.R. Zabaleta, 130 kV, 130A high voltage switching mode power supply for neutral beam plasma heating: design issues, Fusion Eng. Des. 66–68 (2003) 615–620.  A. Nabae, I. Takahashi, H. Akagi, A new neutral-point-clamped PWM inverter, IEEE Trans. Ind. Appl. IA-17 ((5) September/October) (1981) 518–523.  J.R. Coupland, T.S. Green, D.P. Hammond, A.C. Riviere, A study of the ion beam intensity and divergence obtained from a single aperture three electrode extraction system, Rev. Scientif. Inst. 44 (9) (1973) 1258–1270.  V. Toigo et al., Integrated design document of the neutral beam power supply system, Consorzio RFX technical note RFX NBTF TN 75.  V. Toigo, K. Watanabe, Ripple evaluation of acceleration grid voltage, Coordinating Committee on Neutral Beams, Naka, December 12–15, 2006.  M. Bigi, V. Toigo, L. Zanotto, Protection against grid breakdowns in the ITER neutral beam injector power supplies, Fusion Eng. Des. 82 (2007) 905–911.  PSIM User Manual, .  P.K. Steimer, H.E. Gruning, J. Werninger, E. Carroll, S. Klaka, S. Linder, IGCT-a new emerging technology for high power, low cost inverters, in: Proceedings of the 32nd Industry Application Society Annual Meeting, New Orleans, Louisiana, October 5–9, 1997.  A. Ferro, V. Toigo, L. Zanotto, Supporting analyses for the integrated design of the acceleration grid power supply, Consorzio RFX technical note RFX NBTF TN 76.  M. Bigi, A De Lorenzi, L. Grando, K. Watanabe, M. Yamamoto, A model for electrical fast transient analyses of the ITER NBI MAMuG accelerator, this conference.