Electric Power Research Institute Qinghe Beijing 100085 P.R.China Abstract: Commutation failures(CF) are very frequent dynamic events in HVDC system. In the paper, a simplified thyristor lumped-charge model are used to analyzethe change of the stored charges in the device during turn-off. And the commutation process in an HVDC inverter is studied theoretically. The influences of extemal circuit parameters, as well as, thyristor device parameters on CF, is examined. Based on the removal process of the stored charges in thyristor devices, the duration of the phases in tum-off process of the deviceis given, and a novel criterion of commutation failure is proposed. Key Words: Commutation failure Thyristor valve Lumped- charge model Reverse recovq process I. INTRODUCTION Commutation failures are very frequent dynamic events in HVDC system. Most commutation failures are caused by AC bus voltage disturbances due to AC system faults. However, in commutation process, the key component of an HVDC inverter is the thyistor valves. In fact, a current commutation process inan inverter is at least associated with two thyristor valves. When one valveis triggered on, the other is going to turn off. For the turning off valve, the intemal excess charges which are mainly stored in the base zoneof thyristor devices during forward conduction must be removed before it can establish a forward voltage blocking capability. lhis is called reverse recovery effect. If forward voltage is imposed across the turning-off device before the storage charges in thedevice have been removed completely, the thyistor device cannot turn off successhlly. And as a result, CF happens. Early studies onCF were always based on simplified inverter circuit without considering the behavior of thyristor valves during commutation process. And the reverse recovery characteristic of valves are only assumed as a constant, i.e., margin angle[l]. The report of GIGRE WG 14.02 provided another method to simulate CF process in which varying negative voltage-time area represents the reverse recovery characteristic of valves. However, the effects of the parameters of thyristor valves during commutation process are not investigated intensively[2]. In the paper, a simplified thyristor lumped-charge model are used to study the commutation process of an HVDC mverter. And the espressions of the stored charges and the reduction rate of the charges in turning-off thyristor valve during commutation process are derived. Fromthese espressions, the influences of the circuit parameters, as well as, thyristor device parameters, is examined. 11.ASSUMPTI ONS To analyzethe commutation process in an HVDC inverter theoretically, the following assumptions have to be made. (a)Commutation circuit is simplified as Fig.1, where ea, C, and e, represent the AC side phase-voltage respectively. L, is the commutating inductance of the rk-emal circuit. LI represents the DC side current. i l , i5 is the commutating current flowing through valve 1 and valve 5 reqectively. DC current LI nil1 be commutated fromvalve 5 to valve 1, (b)Thyristors in a valve, such as VI, V5, V6, have the samecharacteristics, thus the valve can be regarded as a single thyistor device, (c)Neglect the voltage drop across the valve. before the valve withstands the reverse voltage, (d)Neglect the forward tum-on transient process of valve 1, (e)There is an ideal DC current controller in the DC side of the HVDC system, i.e. LJ remains.constants d ~ g the commutation process. For a 3-phase ground fault in A side, current commutation starts at mt=a shown in Fig. 2,where a is the triggering angle of valve 1. - 503 - A Fig. 1 Shplified commutation circuit -!: P 4 Fig. 2 Commutation process between valve 1 and valve5 111. SIMPLIFIED MODEL OF THYRISTOR It is well known that commutation process in an inverter is directly affected by the reverse recovery process of the thyristor valves used in the inverter. Generally the reverse recovery of a thyristor can be simulated accurately by solvtng a set of semiconductor physical equations using computer. But it is L'ev hard to get an analytic result. Here some simplifications have been made to model thyristor by lumped-charge method[3] and the commutation process is studied theoretically. As stated above, valve 5 in Fig.2 can beconsidered as a single device. Its structure and the carrier distribution in on state are showninFig.3. Theouter region ,PI and NZ are generally heavily doped. The inner PZ region has a higher doping concentration than the NI base region. Assumethat during on state, the total escess charse is stored in the N-baseand located in the middle of this zone. Because the lifetune of carrier in N-baseis larger than the lifetime in P-base, and N-baseis rvider thanP-base, the assumption is valid. (1) QE =@(PE - P,e) O B =&''(PB - P j O ) where, QE is the storage charge at the border of N-basezone, QB is the storage charge in N-basezone, q is the unit electron charge, A is the junction area, pl and p2 are the average hole concentrations in the region corresponding to QE and QB respectively, p is the equilibrium hole concentration, 6 and d are the width of the two charge storage regions respectively(&<d). K S - H - - - - - - d d l , Fig. 3 Structure and carrier distribution of thyristor(on state) According to the ambipolar diffision equation (ADE), the anodecurrent i A which flows in NI can beexpressed as, a.here, D., is the ambipolar diffusion constant, 0, E qp,rL4, T =d' / ( 4Do) s Q,, = T - K - I ~ + Q ~ (4 The charge control continuity equation for QB is Clt t - AI =K . i., whereAI is thenet partlcle current floning into the region times the'electromc charge q, 7 IS the lifetime of minority carrier in N base. R is the gain ofn-pn setion ofthe device. Dung the reverse recovery, Qo and QB decrease. OnceQa reduces to qAdp,o, the space charge region (SCR) appertrs at the left side of the N base zone. With the increase of the width of SCR, the de\<ce will withstand the reverse voltage and the reverse current will decrease (as shown in Fig. 4). In general, SCR at the n&t side of N base is always developed later and slower than SCR at the left side. This means that almost all the voltagedrops on the left SCR ( V A E ~) during reverse recoveq process. Refer to the work of Benda[4], the voltage drop on the left SCR can be given by the following equations., c h u , = L . - . t Eli, (7) dt /l" -+pp 411 where wsc,is the width of the left SCR., p,,, pp are the drift rate of electron and hole respectively in N-base, n =- - I,.? 4 - 504- I V. ANALYSIS OF COMMUTATION PROCESS According to the waveformof reverse recovery current of thyistor device shown in Fig.4, commutation in an HVDC inverter can bedivided into three phases. Phase 1: a / U <t S t o. In this phase, the anodecurrent of valve5 reduces fromLJ to zero. The following equations can be gotten fromthe commutation circuit shown in Fig. 1. where, E is the RMS value of AC line voltage. i, =I ,--E(cosa-cosmt) =wLr . (1 1) Jz 2-E; To thyristor valve5, following equation can be derived from (j )and(l l ). dQB O B - f i -+--K[ I ,--E(cosa-coswf)](12) dt z 2X" In addition, before current commutation, i.e. wf I a , Q, 3 K51d , this is the initial condition of QB. Assume that the value of E keeps constant during commutation process, the storage charge in the base zone of valve5 can be solved as, 45 I - - -_ 0, =C. e +K( I d --Ecosa)r 2- Y" (13) J z 1 1 2x7 T +K- E(-cos wf +w sinwt) .~ 1 (-y +U2 5 1 -cos a i w si na where C =K-E(rcosa - ' 1 . ,+CO- 5- 1 43 2 .\; Generally, x 2 >>w' , then (1 3) can besimplified as , Jz 2x7 Q, =Kr[I, --E(cosa-cosai)] (14) J z - . t-' 2Xr 5 +K- ET',ZU[SIII~ - Sha .eSp(-L)] And the reduction rate of QB can be given as, Jz t-' 2x, 5 - K - Erw[sin wt - si na . e.\p( -2)] Fig.4 Waveformof the reverse recovery current of valve5 Becauseinthisphase 7~/ 2<a<wt <l r , cosot<@and s i nwt - s i na- e~~( - ~) >O . From (l j ), it is shown dQ,/dt <0 . This means that the chargein the base zoneof valve5 is reducing. And its reduction rate is proportional to E. Therefore the edema1 AC side voltage directly afFects the reduction of the storagecharge in base zoneof thyristor valves. On the other hand, the lifetime of the minority carriers in base zone determines thequantity of the storage charge. For ij(fO)=O. the interval of the first phaseof commutation, Ato becomes: t - " r JZI s Afo =[cos-'(cosa - 2) - a] / w (16) E Phase 2: To <t I t, . This phase is fromis=O to 0~0. In this phase the voltagedrop of thyristor valve5 can also beneglected, thus the expressions of (20 and 0s is the sameas thosein phase 1. For i,,,.,, =0. (17) f i 2x7 I , - - E(cos a - coswt, ) =0 For Q,,,;,,, =0 ' (19) And i j ( t l ) = I , --E(cosa-coswt,) 2Xr where I , =t o +At, . From( 17), ( 18) and ( 19), the following J2 equation can bederived. [w cos ot, - (1 +T) -sin wfo]Att, T 1 . b5 T (20) At + At +sin ut, - sin a . e xp( - -) =0 T where sin(wAt, ) 2 W. At, , cos(o&, ) 1 are assumed. It is shown fromcalculation that when At, <5 OOj a , the relative -505- error of this assumption is below 0.4 1%. Because the lifetime of the minority carriers in the base zone are very small (about severaldecade microseconds), and Ato +At, is about several milliseconds which is several decades of 7, the third part of the left side in (20) can be neglected. Thus the duration of the second phase, Atl, can be derived. b (21) sin mt , (1+-)sinwt, -rwcoswt, T Kr At, = Phase 3: tl<t<tz. It can befound fromFig.4 that 00 =U when -1. After this, thyristor valve 5 begins to withstand reverse voltage and its anode current is starts to decrease. Once is reaches b FF, the forward blocking capacity of the device will be recovered. In this phase the anode-cathode voltage vx of valve 5 cannot be neglected. And the equations of the extemal circuit are modified as, di, dii dt dt L. - - L. --I- vAK =euc It is obvious that QB can be solved by digital simulation methods such as Ronga-Kutta method, kom(j), (8) and (9). However the analytical result cannot be gotten from these equations. In fact, it is found that in this phase the current of valve 5 decreases exponentially[5] i.e., 1 4 , __ i = - I J e 2 (23) where I , <<I , , . Imif is the peak value of the reverse recovery current. IOFF is the current value below which the thyistor device turnof successfully. Because O,( tl ) =0 , the following equation can begotten from(4), QB( rl ) =- T- i 5( t , ) z T-I,,, (25) Theabove equation shows that at the moment tl the storage charges left in the base zone of valve 5 is T . I , . Once is reduces to bm, the excessive charges in the base zone are removed completely. The transmission time T can be derived fiom(24) and (25). T = A ( I - L ) I (26) I , The duration of the third phase At2which is the decrease time of i5 fromI mf to IOFF i s shown as, At2 =/Z.I nLzT.lnL (27) I*,, Thus from the moment &=a at which valve 1 is triggered on and the commutation begins, to the moment at - 506 - which valve 5 tums off completely, the duration of the whole process At is the sumof the duration of the three phases. So the criterion of commutation failure is : At 2 p/ w . It can be Ar =Ato -k At, +Atl (28) known form(16), (21) and (27) that Ab arerelated with the parameters of the extemal circuit. And AtI, At2 arenot only affected by the parameters of the extemal circuit but also associated with the micro parameters of thyristor devices. Therefor the criterion of CF should be considered both the .external circuit and the micro process of thyistor device. V. CONCLUSION Using a simplified lumped-charge model of thyristor, commutation process of an HVDC inverter is investigated. It was found that not on!y the AC commutation voltage disturbance but also the parameters of the thyristor valves affect the commutation process. The lifetlme of the mlnority camers in the thyristor is a very influencing factor to the stored charges in the device. Based on the removal process of the stored charges in thyristor devices, the duration of the phases in the turn-off process of the device is given, and a novel criterion of commutation failure is proposed which may be used in EMTF. Because the criterion is derived with considering of both the extemal circuit conditions and the inner charge removal process in thyristor device, it is more accurate than the conventional one. VI. ACKNOWLEDGES The authors would like to thank Chinese National Nature Science Foundation(NNSF), Technology Department of Electric Power Ministry and Northeast China Electric Power Group Inc. for the financial support ofthis work. VII. REFERENCES [ 11 Zhejiang University, DC Transmission System (Chinese), Hydraulic and Electric Power Press, 1985.6. [2] WG 14.02 Commutation Failures in HVDC Transmission Systems Due to AC SystemFaults, Electra. No.169, Dec. 1996 [3] C.L.Ma et al., A physically-based lumped charge SCR model, IEEE Power Electronics Specialists Cod., Seattle, 1993 [4] H.Benda, E.Spenke, Reverse recovery processes in Silicon power rectifiers , Proc. IEEE, Vo1.55, No.8, pp.1331-1354, 1967 [S]C.W.LEE,S.B.Park, %sign of a thyristor snubber circuit by considering the reverse recovery process IEEE Trans. Power Electron. Vo1.3, No.4, pp.440446, 1988
IEEE Transactions On Power Electronics Volume 6 Issue 2 1991 (Doi 10.1109/63.76804) Lauritzen, P.O. Ma, C.L. - A Simple Diode Model With Reverse Recovery