scieee Open visual document viewer

Real time control developments at JET in preparation for deuterium-tritium operation

Lennholm, M.; Carvalho, I.S.; Cave-Ayland, K.; Chargnard, A.; Challis, C.; Jet Contributors; García Muñoz, Manuel

Abstract

Robust high performance plasma scenarios are being developed to exploit the unique capability of JETto operate with Tritium and Deuterium. In this context, real time control schemes are used to guidethe plasma into the desired state and maintain it there. Other real time schemes detect undesirablebehaviour and trigger appropriate actions to assure the best experimental results without unnecessaryuse of the limited neutron and Tritium budget. This paper discusses continuously active controllers andevent/threshold detection algorithms triggering a variety of actions. Recent advances include: (i) Con-trol of the degree of plasma detachment via impurity injection; (ii) ELM frequency control via gas/Pelletinjection; (iii) Sawtooth pacing using ICRH modulation, (iv) control of the Hydrogen to Deuterium iso-tope ratio through gas injection and (v) the determination that a discharge is not evolving as desired,triggering a cascade of actions attempting to stop the plasma rapidly and safely, eventually triggering massive gas injection if a disruption is deemed unavoidable. For high power Deuterium-Tritium operation these control schemes need to be integrated into the plasma scenarios ensuring that they are mutually compatible,

Full text

Fusion Enginee ing and Design 123 (2017) 535–540 Con en s lis s a ailable a ScienceDi ec Fusion Enginee ing and Design jo u nal home p age: www.else ie .com/loca e/ usengdes Real ime con ol de elopmen s a JET in p epa a ion o deu e ium- i ium ope a ion M. Lennholma,b,k,∗, I.S. Ca alhoc,d,k, K. Ca e-Aylandd,k, A. Chagna de,k, C. Challisd,k, R. Fel ond,k, D. F igione ,k, L. Ga zo id,k, A. Goodyea d,k, J.P. G a esg,k, C. Guillemau c,d,k, J.R. Ha isond,k, E. Le ched,h,k, P.J. Lomasd,k, R. Mooneyd,k, F. Riminid,k, A.C.C. Sipsa,b,k, C. Sozzii,k, D. Valca celd,k, J. Vegaj,k, JET con ibu o s1 aEu opean Commission, B-1049 B ussels, Belgium bJET Exploi a ion Uni , Culham Science Cen e, Abingdon OX14 3DB, UK cIns i u o de Plasmas e Fusão Nuclea , Ins i u o Supe io Técnico, Uni e sidade de Lisboa, P-1049-001 Lisboa, Po ugal dCCFE, Culham Science Cen e, Abingdon OX14 3DB, UK eEcole Cen ale de Lille, 59651 Villeneu e d’Ascq, F ance ENEA, C.R. F asca i, Roma, I aly gEPFL, CRPP, CH-1015 Lausanne, Swi ze land hLabo a o y o Plasma Physics, Koninklijke Mili ai e School, Ecole Royale Mili ai e Renaissancelaan, 30 A enue de la Renaissance, B-1000 B ussels, Belgium iIs i u o di Fisica del Plasma, CNR, Milano, I aly jLabo a o io Nacional de Fusión, CIEMAT, Mad id, Spain, Spain kEURO usion Conso ium, JET, Culham Science Cen e, Abingdon OX14 3DB, UK h i g h l i g h s •Real ime con ol schemes ha e been de eloped o op imise JET pe o mance in DT wi h ITER like wall. •De achmen con ol ia impu i y injec ion. •ELM equency con ol ia gas/Pelle injec ion. •Saw oo h pacing using ICRH. a i c l e i n o A icle his o y: Recei ed 29 Sep embe 2016 Recei ed in e ised o m 3 Feb ua y 2017 Accep ed 5 May 2017 A ailable online 17 May 2017 Keywo ds: Real ime con ol T i ium ope a ion Plasma e mina ion ELMs Saw ee h De achmen Iso ope con ol a b s a c Robus high pe o mance plasma scena ios a e being de eloped o exploi he unique capabili y o JET o ope a e wi h T i ium and Deu e ium. In his con ex , eal ime con ol schemes a e used o guide he plasma in o he desi ed s a e and main ain i he e. O he eal ime schemes de ec undesi able beha iou and igge app op ia e ac ions o assu e he bes expe imen al esul s wi hou unnecessa y use o he limi ed neu on and T i ium budge . This pape discusses con inuously ac i e con olle s and e en / h eshold de ec ion algo i hms igge ing a a ie y o ac ions. Recen ad ances include: (i) Con- ol o he deg ee o plasma de achmen ia impu i y injec ion; (ii) ELM equency con ol ia gas/Pelle injec ion; (iii) Saw oo h pacing using ICRH modula ion, (i ) con ol o he Hyd ogen o Deu e ium iso- ope a io h ough gas injec ion and ( ) he de e mina ion ha a discha ge is no e ol ing as desi ed, igge ing a cascade o ac ions a emp ing o s op he plasma apidly and sa ely, e en ually igge ing massi e gas injec ion i a dis up ion is deemed una oidable. Fo high powe Deu e ium-T i ium ope a ion hese con ol schemes need o be in eg a ed in o he plasma scena ios ensu ing ha hey a e mu ually compa ible. © 2017 The Au ho s. Published by Else ie B.V. This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/). ∗Co esponding au ho a : JET Exploi a ion Uni , Culham Science Cen e, Abing- don OX14 3DB, UK. E-mail add ess: [email p o ec ed] (M. Lennholm). 1See he Appendix o F. Romanelli e al., P oceedings o he 25 h IAEA Fusion Ene gy Con e ence 2014, Sain Pe e sbu g, Russia. 1. In oduc ion The JET okamak is he only cu en ly ope a ing okamak which can ope a e wi h he Deu e ium-T i ium uel mix equi ed in a nuclea usion eac o . The cu en JET plan en isages ope a ion wi h pu e T i ium (TT) plasma in 2018 ollowed by Deu e ium- h p://dx.doi.o g/10.1016/j. usengdes.2017.05.023 0920-3796/© 2017 The Au ho s. Published by Else ie B.V. This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4. 0/). 536 M. Lennholm e al. / Fusion Enginee ing and Design 123 (2017) 535–540 Fig. 1. (a) A simple closed loop algo i hm wi h compensa ion o s a ic nonlinea - i y. (b) Schema ic illus a ion o he implemen a ion o his con olle in he RTCC language. T i ium (DT) ope a ion in 2019. Gaining u he expe ience in ope a ing a nuclea okamak is conside ed essen ial o he suc- cess o ITER and he cu en and u u e JET p og amme is s ongly ocused on ga he ing he maximum in o ma ion and expe ience om he upcoming TT and DT campaigns. The in o ma ion sough encompasses he iso ope scaling o H-mode h eshold, confine- men , ELM equency e c. bes ga he ed h ough TT expe imen s. The DT ope a ion will p o ide aluable in o ma ion on he physics o a plasma wi h significan alpha pa icle hea ing and neu on p o- duc ion. JET has unde gone majo changes since he p e ious DT expe imen s in 1997 [1], he mos impo an o which is he ins al- la ion o he all me al ITER like wall in 2010–11, using Be ylium as fi s wall ma e ial in he main chambe and Tungs en in he Di e - o , as o eseen o ITER [2–4]. Ope a ion wi h DT uel and wi h he ITER like wall poses a se ies o specific challenges as desc ibed in Sec ion 3. The cu en pape in es iga es he ole ha eal ime con ol can play in mee ing hese challenges. 2. JET eal ime con ol a chi ec u e Two co e con ol sys ems a e equi ed o he ou ine ope a ion o he JET machine: The Plasma Posi ion and Cu en Con ol sys em (PPCC) uses he poloidal coils as ac ua o s while he plasma den- si y eedback con ol sys em (PDF) uses gas injec ion as i s ac ua o . Though hese sys ems a e no discussed u he , i is unde s ood ha hey a e ac i e oge he wi h all he con olle s discussed. The con olle s desc ibed in he ollowing a e unning in a cen al con- olle (RTCC) which ecei es da a, in eal ime, om a la ge numbe o JET measu emen sys ems and eal ime p ocesso s [5]. RTCC can ou pu eques signals o a ious ac ua o s, no ably hea ing and gas in oduc ion sys ems bu also o PPCC and PDF. The p og amming o con olle s in RTCC is done in a high le el, block diag am language whe e each block in a s anda d block diag am is ansla ed in o a line in he p og amme. The use can p og am a con ol algo i hm wi hou ha ing o wo y abou in e ace issues and as a conse- quence simple con ol algo i hms can be implemen ed apidly – e en ‘on he fly’ du ing sessions. Fig. 1a shows a simple single inpu single ou pu (SISO) con- ol scheme which is ep esen a i e o a numbe o he con olle s used a p esen a JET. The con olle includes a nonlinea unc ion which co ec s o a known s a ic nonlinea i y in he inpu ou pu map. The simple implemen a ion in he RTCC high le el language is shown schema ically in Fig. 1b. Fo a obus implemen a ion a ew bells and whis les ha e o be added including hings like limi s and an i-windup. Fou sepa a e con ol algo i hms can un simul aneously wi hin RTCC. Mul i inpu mul i ou pu (MIMO) con olle s wi h up o 3 Fig. 2. ‘Dud’ de ec ion. Two discha ges om he 1997 JET Deu e ium-T i ium expe - imen s which we e s opped when he neu on yield no longe exceeded he equi ed minimum. The dashed lines ske ch discha ges planned o he nex JET Deu e ium- T i ium campaign. inpu s and 3 ou pu s ha e been implemen ed in a single algo i hm and his is close o he cu en limi o he sys em capabili ies. The possibili y o upg ading he sys em o allow mo e compli- ca ed algo i hms, while simpli ying he in e ace o he ope a o s, is unde in es iga ion. 3. Ope a ional challenges and he ole o eal ime con ol in mee ing hem 3.1. Tungs en influx The dominan change associa ed wi h he ins alla ion o he ITER like wall on JET, al eady documen ed on ASDEX-Upg ade [6], is he endency o ungs en o accumula e nea he plasma cen e [7,8]. Such accumula ion can lead o a adia ion collapse in a iably caus- ing a dis up ion. Se e al echniques ha e been de eloped o a oid o coun e his accumula ion. I u ns ou ha ELMs, while spu e - ing ungs en om he di e o , also help o expel ungs en om he plasma edge. As a consequence a high ELM equency can help limi impu i y influx [9]. Once he ungs en has mo ed beyond he each o he ELMs i mo es owa ds he cen e o he plasma. He e saw- ee h a e seen o expel ungs en e ec i ely owa ds he ou e pa o he plasma [10,11]. A final ool, e ec i e in p e en ing exces- si e ungs en accumula ion, is cen al elec on hea ing [12]. On JET such hea ing can be p o ided by ICRH, whe eas ECRH has been shown o be e ec i e on ASDEX-Upg ade. E ec i e a oidance o ungs en accumula ion using hese ools is eadily achie able on JET, bu a hea y p ice can be paid in e ms o confinemen i ELM and saw oo h equencies a e inc eased excessi ely [13–15]. Real ime con ol o ELM and saw oo h equencies can play an impo - an ole in achie ing he op imal comp omise, a oiding ungs en accumula ion while main aining good confinemen . 3.2. Di e o ene gy handling A second, hough linked, challenge posed by he me al wall is i s limi ed powe and ene gy handling. Main aining a su ficien ly high ELM equency is again desi able, hough o he echniques also need o be employed. These echniques include sweeping he di e - o s ike-poin loca ion and he in oduc ion o ligh impu i ies in o he di e o egion o adia e ene gy locally, c ea ing a pa ially o ully de ached plasma. Finally a educ ion in he inpu powe may p o e necessa y. Real ime con ol can play a c ucial oll in con- M. Lennholm e al. / Fusion Enginee ing and Design 123 (2017) 535–540 537 Fig. 3. Saw oo h and N = 2 ac i i y o wo discha ges: ‘(a) Cons an ICRH powe . Long saw ee h igge N = 2 ea ing modes. (90142); (b) E ec i e saw oo h pacing leading o he a oidance o he N = 2 modes. olling adia ion, de achmen and di e o hea load allowing he bes possible plasma pe o mance while s aying wi hin he limi s imposed by he di e o powe and ene gy handling capabili y. 3.3. Dis up ions A less ob ious complica ion in he ope a ion wi h he me al wall is he ac ha dis up ion o ces ha e inc eased signifi- can ly in compa ison wi h ca bon wall ope a ion [16,17]. This is exace ba ed by he be yllium walls in he main chambe being p one o flash mel ing du ing dis up ions. P e en ion, p edic ion and mi iga ion o dis up ions ha e he e o e become e en mo e impo an . A oiding ungs en accumula ion is p obably he mos impo an p e en ion ac ion, hough p e en ing he igge ing o NTMs h ough saw oo h sho ening and limi a ion o be a a e also impo an . Ea ly p edic ion ha he plasma is heading owa ds a dis up ion may allow he dis up ion o be p e en ed by ini ia ing a p ede e mined e mina ion scena io [18,19]. A ange o e mina- ion scena ios ha e been de eloped, each op imised o he specific condi ions igge ing he ini ia ion o he e mina ion. Once i is clea ha a dis up ion is imminen he main mi iga ing ac ion a JET is he fi ing o he dis up ion mi iga ion al e injec ing a la ge amoun o gas e mina ing he discha ge apidly while adia ing mos o he plasma ene gy. 3.4. Dud de ec ion The main conce n when ope a ing wi h i ium, especially in ac i e ope a ion wi h deu e ium- i ium plasmas, is o make op i- mal use o he limi ed amoun o i ium a ailable and o consume he se e ely es ic ed neu on budge wisely. De e mining, in eal ime, when a discha ge is unlikely o be o scien ific alue and, i so, e mina ing i sa ely can esul in significan educ ion in neu on p oduc ion and T i ium consump ion. In DT discha ges he simples way o do his, al eady exploi ed du ing he JET DTE 1 expe imen s in 1997, is o moni o whe he he neu on a e emains abo e a p ede e mined cu e. Fig. 2 shows wo ho ion ELM ee H-mode discha ges om 1997 when his me hod was used [1]. The bes o hese wo pulses eached a peak usion powe o ∼12 MW. E en o he sho pulses in ques ion a significan neu on sa ing was achie ed by s opping one pulse ea ly. No e ha he second pulse was also s opped by he sys em, bu only a e he gian ELM which always e mina ed he high usion yield phase o hese discha ges. Fig. 2 also includes a p ojec ion o he mo e s eady discha ges planned o he upcoming deu e ium- i ium campaign, showing ha a la ge neu on sa ing can be ob ained by s opping ‘bad’ pulses ea ly. In he planned discha ges highe NBI powe , compa ed o 1997, will be equi ed o achie e he same usion powe due o he need o un wi h egula ELMs. O he signals, such as impu i y con- en and hea ing powe can also be used o de e mine whe he a discha ge should be e mina ed ea ly. A he de ec ion o a ‘dud’ a e mina ion simila o he no mal end o a heal hy pulse is ins i- ga ed. Du ing such a e mina ion, o he e en s may ins iga e mo e d aconian ac ions in a p og essing hie a chy o se e i y. An exam- ple could be he de ec ion o wall o e hea ing igge ing a change o configu a ion. La e ; he de ec ion o an inc eased isk o dis up- ion may igge p e en ion and mi iga ions ac ions as desc ibed abo e. 3.5. Iso ope con ol A second issue, which will be mo e impo an du ing DT ope - a ion, is he con ol o he iso ope a io as equi ed o achie e maximum usion powe . Con olle s de eloped o his pu pose a e closely ela ed o exis ing mino i y concen a ion con olle s used in ICRH hea ed discha ges. 4. Real ime con olle s in use a JET A numbe o eal ime con olle s ha e been de eloped a JET, each con ibu ing o mee ing one o mo e o he challenges ou lined abo e. In he ollowing, expe imen al e idence o he e ec i eness o each o hese con olle s is discussed. 4.1. ELM equency con ol An ELM equency con olle is ou inely used a JET. This con- olle exploi s he ac ha , unde mos condi ions, he ELM equency inc eases wi h gas injec ion a e. This con olle , e ec- i e in i s own igh , is also egula ly used as a ‘sa e y ne ’, only ac ing when he ELM equency d ops below a h eshold. When his happens he con olle injec s gas o main ain he equency a he equi ed minimum. When igge ing ELMs h ough pelle injec- ion, his is pa icula ly use ul due o he a iabili y in he pelle ELM igge ing e ficiency [13,20]. 4.2. Saw oo h pacing Con olling he saw oo h pe iod wi h a iew o p e en ing la ge saw oo h c ashes om igge ing o Neoclassical Tea ing Modes (NTMs) has been in es iga ed on JET o e he las decade [21–24]. All he echniques de eloped in his esea ch ely on loca ing he 538 M. Lennholm e al. / Fusion Enginee ing and Design 123 (2017) 535–540 Fig. 4. (a) Ske ch showing a achmen ac ion dA = Isa /I oll as a unc ion o inpu gas flow a e (blue) and an exponen ial cu e ma ching his o ni ogen a es beyond he ollo e ( ed). (b) and (c) Closed loop de achmen con ol wi hou (b) and wi h (c) compensa ion o he non-linea i y seen in (a). No e ha in he discha ges in (c) he eedback was ac i a ed when he a achmen ac ion eached 0.8 as de e mined h ough p io knowledge o I oll. (Fo in e p e a ion o he e e ences o colo in his figu e legend, he eade is e e ed o he web e sion o his a icle.) Fig. 5. (a) Hyd ogen-Deu e ium mix u e con ol. (#91234). The con olle compensa es o he dis u bance in oduced by a ia ions in he deu e ium neu al beam powe . The al e openings a e limi ed by he con olle o 20% o ‘ ully open’. (b) Simul aneous con ol o he ELM equency and He3concen a ion by injec ion o Deu e ium and He3gas. The hi d ace shows he beha iou o he ‘linea ized equency’ achie ed by applying −1 o bo h he eques ed and measu ed equency as illus a ed in Fig. 1(c) Simul aneous ope a ion o a no malised be a con olle by NBI and a bang/bang con ol scheme aimed a eac ing o adia ion peaking. A 12.2 s a discha ge e mina ion is igge ed due o he de ec ion o MHD ac i i y. ICRH deposi ion nea he q = 1 su ace o des abilise he m,n = 1,1 mode, inducing mo e equen saw ee h. Simila echniques using ECRH ha e been explo ed on o he Tokamaks [25–28]. TCV has also demons a ed he pacing o saw ee h using modula ion o cen ally deposi ed ECRH [29,30]. Gi en he s ong s abilising e ec o cen- al ICRH, pacing saw ee h using modula ed cen ally deposi ed ICRH should be e y e ec i e [31]. This has indeed been p o en by ecen expe imen s on JET [32]. In hese expe imen s he ICRH modula ion was con olled in eal ime, swi ching he ICRH powe o when he ime since he p e ious saw oo h c ash exceeded a h eshold and swi ching i back on when a new saw oo h c ash is de ec ed. Fig. 3 shows wo pulses, (a) wi hou saw oo h pacing and (b) wi h saw oo h pacing. In (a) long saw ee h lead o he igge - ing o N = 2 modes (p obably weak ea ing modes) while saw ee h a e e ec i ely paced in (b) leading o he comple e absence o long saw ee h and N = 2 modes. In (b) he h eshold o swi ching o he ICRH is se o 0.15s. The ime be ween he eques o swi ch o ICRH and he igge ing o a saw oo h is ∼50 ms in his pulse, leading o a saw oo h pe iod <0.2 s h oughou he pulse. No e ha he eal ime con ol means ha he ICRH is no swi ched o when he na u al saw oo h pe iod is su ficien ly sho . The eal ime con- ol he e o e maximises he ICRH du y cycle allowing he op imal use o ICRH o cen al hea ing and e ec i e ungs en sc eening, while assu ing NTM a oidance and ungs en flushing by keeping he saw oo h pe iod low. 4.3. De achmen con ol A eedback con olle , con olling he di e o de achmen ac- ion has ecen ly been de eloped a JET [33]. The con olle uses he ion sa u a ion cu en (Isa ) measu emen s om an a ay o Lang- mui p obes, si ua ed in he di e o , o de e mine he deg ee o de achmen in eal ime. The de achmen is con olled, in eedback, by injec ing ni ogen in o he di e o . Fig. 4a shows a ske ch o he s eady s a e map ela ing inpu gas o ion sa u a ion cu en o he Langmui p obe si ua ed closes o he di e o s ike poin . This cu e exhibi s a maximum and he alue a his poin is e med he ‘ oll o e ’ sa u a ion cu en I oll. The deg ee o de achmen dDe a his poin is ze o by defini ion. A highe impu i y injec ion a es he deg ee o de achmen is defined as dDe =(I oll − Isa )/I oll, which eaches one when he ion sa u a ion cu en d op o ze o. Fo simplici y he con olled alue is he ‘a achmen ac ion’: dA = 1 − dDe = Isa /I oll. Wi h his defini ion Fig. 4a di ec ly ans- la es in o showing a achmen ac ion as a unc ion o gas injec ion M. Lennholm e al. / Fusion Enginee ing and Design 123 (2017) 535–540 539 Fig. 6. MIMO con olle – Top) 3 × 3 con ol block diag am. The decoupling ma ix is he in e se o he s eady s a e plasma ma ix in he linea iza ion poin . Bo on) Equi alen block diag amme, showing how he decoupling, i pe ec , esul s in 3 independen SISO con ol loops. a e. As he inpu ou pu map is non mono onic he con olle sign mus change when mo ing om le o igh o he oll o e poin . This is handled by s a ing wi h a cons an , la ge, gas flow, con inu- ously de ec ing he maximum Isa which has been eached since he s a o he discha ge. Once Isa d ops clea ly below his maximum he con olle assumes ha he igh hand side o he inpu ou pu map, whe e dA dec eases wi h inc eased gas injec ion, has been eached and closed loop con ol is s a ed. Fig. 4b shows he esul o wo simila discha ges, one sho - ened o ope a ional easons, using his con olle wi h di e en eques ed a achmen ac ions bu wi h iden ical con olle se - ings. In he case wi h a la ge eques ed a achmen ac ion, s ong con olle oscilla ions occu . The discha ge wi h a smalle a achmen ac ion eques shows no such oscilla ions. The sho du a ion o his discha ge makes i di ficul o asce ain wi h ce - ain y ha no oscilla ion would de elop in a longe pulse. The obse a ion ha he ni ogen injec ion ac ua o ha dly a ies in his discha ge in sha p con as o he significan a ia ion seen in he ea ly pa o he oscilla ing discha ge, leads us o su mise ha he con olle is s able in he case o low a achmen eques . The ac ha low a achmen ac ion leads o s able con olle beha iou while a highe a achmen ac ion esul s in con olle oscilla ions is caused by he s a ic non-linea i y in he inpu ou pu map. This can eadily be confi med h ough simple con olle simula ions. The non-linea i y can be elimina ed om he closed loop by in o- ducing a non-linea compensa ing block in he con ol diag am as sugges ed in Fig. 1 esul ing in s able ope a ion o e a ange o a achmen ac ions as illus a ed by Fig. 4c. The exponen ial cu e used o p oduce he linea iza ion, alid o ni ogen injec ion la ge han he ollo e alue, is shown schema ically by ull and dashed ed lines in Fig. 4a. 4.4. Plasma composi ion con ol As men ioned in Sec ion 2 i is desi able o con ol he iso ope a io, especially du ing DT ope a ion. Fig. 5a shows an example whe e he iso ope mix is main ained e ficien ly a 50% Hyd ogen, 50% Deu e ium by such a con olle . Two gas injec ion modules, one injec ing deu e ium and one injec ing hyd ogen, a e used simul- aneously a ying he a io o he wo gas injec ion a es, while Fig. 7. MIMO con olle simula ion, including pe iods when one ac ua o ouches a limi . keeping he o al injec ion a e cons an . The iso ope a io is de e - mined in eal ime using spec oscopic signals [34]. Simila con ol schemes a e also used o con ol he He3mino i y ac ion o op i- misa ion o ICRH abso p ion. 5. Mul i inpu mul i ou pu con ol The p e ious sec ion desc ibed a numbe o indi idual single inpu single ou pu con olle s. When mo ing on o Deu e ium T i ium ope a ion i is impo an ha hese con olle s can ope - a e oge he . Expe ience wi h he simul aneous ope a ion o he con olle s equi ed o high powe ILW ope a ion emains e y limi ed. Fig. 5b shows he JET ELM equency and He3con olle s ope a ing oge he . I should be no ed ha he ELM equency con- olle apidly eaches sa u a ion le el and i is no clea whe he he wo con olle s, which a e likely o exhibi a no insignifican coupling, would ha e ope a ed sa is ac o ily i nei he o hem had been unning agains hei limi s. Fig. 5c shows he ope a ion o he JET s anda d be a con olle using NBI powe as ac ua o com- bined wi h a bang/bang con olle which s eps up he gas injec ion s ongly when he plasma adia ion peaking, defined as he a io be ween a cen al and an o axis bolome e channel, is seen o exceed a ce ain h eshold and educes he gas injec ion again when he peaking e u ns below ano he , lowe , h eshold. In he dis- cha ge shown he gas helps o keep he plasma ali e o a while bu , e en ually, s ong MHD ac i i y igge s a plasma e mina ion. The combina ion o con olle s in Fig. 5 only sc a ches he su - ace o he likely u u e equi emen o combining con olle s. A s udy in es iga ing he simul aneous use o deu e ium injec ion Dinj, NBI powe PNBI and ni ogen injec ion Ninj o con ol ELM equency ELM, no malised be a ˇNand adia ed ac ion FRad, has been unde aken and a decoupled MIMO con olle has been de el- oped. The p inciple used in de eloping his con olle is based on he obse a ion ha he main ime cons an s in his sys em can be associa ed wi h he indi idual ac ua o s. A ound a ce ain ope a - ing poin his allows us o desc ibe he plasma as a 3 × 3 ma ix Mp:  ELM ˇN FRad = Mp·Dinj PNBI Ninj . By inse ing M−1 pin o he con olle , as illus a ed in Fig. 6 ( op), we can elimina e he c oss coupling e ms. Fig. 6 (bo om) shows ha his is equi alen o con olling he linea combina ions ya yb yc= M−1 p · ELM ˇN FRad  owa ds he e - 540 M. Lennholm e al. / Fusion Enginee ing and Design 123 (2017) 535–540 e ences  a b c= M−1 p · ELM Reques ˇN Reques FRad Reques in h ee independen SISO con ol loops. Fig. 7 shows a simula ion o he beha iou o such a con olle . Though he ma ix used o de e mining he decoupling ma ix is based on a specific ope a ing poin , he model used o he simula ion does ake in o accoun , albei c udely, he nonlinea plasma esponse. The simula ion also akes in o accoun ha he NBI powe can only be a ied in s eps o ∼1 MW and i handles he case whe e an ac ua o eached a limi . In his case he con ol a i- able mos closely associa ed wi h he limi ing ac ua o is emo ed and he con olle becomes a 2 × 2 con olle un il such ime ha he ull con olle would eques he limi ing ac ua o o mo e back o he limi . 6. Conclusions Ope a ion o JET wi h he all me al ITER like wall oge he wi h he planned ope a ion wi h Deu e ium T i ium plasma poses a a i- e y o challenges. A numbe o eal ime con olle s ha e been de eloped o help mee ing hese challenges, including ELM and saw oo h equency con olle s, de achmen con olle s and mix- u e con olle s. Assu ing ha hese con olle s can be combined e ec i ely emains he main eal ime con ol ask o be comple ed in p epa a ion o he upcoming Deu e ium T i ium expe imen al campaigns. Acknowledgmen s This wo k has been ca ied ou wi hin he amewo k o he EURO usion Conso ium and has ecei ed unding om he Eu a om esea ch and aining p og amme 2014–2018 unde g an ag eemen No. 633053. The iews and opinions exp essed he ein do no necessa ily eflec hose o he Eu opean Commission. The wo k o A. Chagna d ecei ed he suppo o he F ench Embassy in London. Re e ences [1] M. Keilhacke , e al., Nucl. Fusion 39 (1999) 209. [2] F. Romanelli, e al., Nucl. Fusion 53 (2013) 104002. [3] G.F. Ma hews, e al., Phys. Sc . (2011) 014001. [4] R. Neu, e al., Phys. Plasmas 20 (2013) 056111. [5] Q. King, H. B elen, JET p ep in s JET-P(98) (1998). h p://www.eu o- usionscipub.o g/wp-con en /uploads/2014/11/JETP980241.pd . [6] R. Neu, e al., Plasma Phys. Con ol. Fusion 49 (12B) (2007) B59–B70. [7] E. Jo in, e al., Nucl. Fusion 54 (2014) 013011. [8] M. Beu skens, e al., Plasma Phys. Con ol. Fusion 55 (2013) 124043. [9] T. Pü e ich, e al., Plasma Phys. Con ol. Fusion 55 (2013) 124036. [10] J. G a es, e al., Plasma Phys. Con ol. Fusion 57 (2015) 014033. [11] I. Chapman, e al., Nucl. Fusion 50 (2010) 102001. [12] E. Le che, e al., Nucl. Fusion 56 (2016) 036022. [13] M. Lennholm, e al., Nucl. Fusion 56 (2016) 016008. [14] G. Saibene, e al., JNM 241–243 (476) (1997). [15] M.N.A. Beu skens, e al., Nucl. Fusion 54 (2014) 043001. [16] P.C. de V ies, e al., Phys. Plasmas 21 (2014) 056101. [17] C. Reux, e al., Fusion Eng. Des. 88 (2014) 1101. [18] J. Vega, e al., Fusion Eng. Des. 88 (6–8) (2014) 1228. [19] R. Mo eno, J. Vega, e al., Fusion Sci. Technol. 69 (2016) 485. [20] M. Lennholm, e al., Nucl. Fusion 55 (2015) 063004. [21] L.-G. E iksson, e al., Phys. Re . Le . 92 (2004) 235004. [22] J.P. G a es, e al., Na . Commun. 3 (2012) 626. [23] J.P. G a es, e al., Phys. Re . Le . 102 (2009) 065005. [24] M. Lennholm, e al., Nucl. Fusion 51 (2011) 073032. [25] C. Angioni, e al., Nucl. Fusion 43 (2003) 455. [26] A. Mück, e al., Plasma Phys. Con ol. Fusion 4 (3) (2005) 1633. [27] J.I. Paley, e al., Plasma Phys. Con ol. Fusion 51 (2009) 055010. [28] M. Lennholm, e al., Phys. Re . Le . 102 (2009) 115004. [29] T.P. Goodman, e al., Phys. Re . Le . 106 (2011) 245002. [30] M. Lau e , e al., Nucl. Fusion 52 (2012) 062002. [31] M. Lau e , e al., Plasma Phys. Con ol. Fusion 58 (2016) 124004. [32] E. Le che, e al., Nucl. Fusion 57 (2017) 036027. [33] C. Guillemau , e al., Submi ed PPCF, Plasma Phys. Con ol. Fusion 59 (2017) 045001. [34] D. Valca cel e al., in p epa a ion.