Annals of Nuclear Energy, eel . 3, pp. 359 to 366. Pcr~_ mou Pros 19"/6. Printed in Nortlmm Ireland

FUEL MANAGEMENT IN CANDU REACTORS

C. H. MILLAR Atomic Energy of Canada Limited

Sammary----CANDU fuel management differs completely from that in LWR's. On-power refuelling of the pressurized fuel channels with short (50 cm) fuel bundles is a continuing function of reactor operations and leads rapidly to equilibrium core conditions, refuelling rate, and fuel burn-up. Single-bundle fuel shifts in the 25 MWe, low-rated NPD reactor have been replaced with multi- bundle shifts in the 540 MWe reactors at Pickering to avoid excessive fuelling machine operation and fuel failures caused by moving irradiated fuel to higher flux positions. "Principles of fuel manage- ment in Pickering are described together with the operational experience on which these principles are based. Fuel management problems for other versions of the CANDU reactor (boiling light water coolant, Pu-recycle, Th/U-233 cycle) are outlined briefly.

B A S I C C O N C E P T

Fuel management in CANDU, the Canadian heavy water, natural uranium, pressure tube reactor, is completely different from that in enriched light water reactors. The original concept of fuel manage- ment was to refuel one of the short (0.5 m) fuel bundles at a time by means of a pair of fuelling machines that could be attached to the ends of the horizontal fuel channel while the reactor was at full power. A fresh bundle was inserted from the machine at one end of the channel, while a spent bundle was ejected into the machine at the other end. This system avoided the necessity for refuelling shutdowns and for building excess reactivity into the reactor together with the corresponding shim rod capacity. It also led to uniformly high average burnup in all fuel material. It required, however, that refuelling be a continuous part of reactor operations.

A second aspect of the concept was to move the fuel through alternate channels in opposite directions so that the fuel near both reactor ends was an equal mixture of fresh and nearly-spent fuel, and thus of approximately the same average reactivity as the half-spent fuel at the centre. This bi-directional fuelling--with the concomitant bi-directional coolant flow--resulted in uniform average lattice characteristics throughout the reactor.

F U E L M A N A G E M E N T I N N P D

Fuel management schemes for the 25 MWe Nuclear Power Demonstration reactor started up in 1962 were first studied by Barss (1962) who showed that once the initial fresh-fuel reactivity transient had been overcome (in 6 months to a year of operation) the equilibrium fuelling rate was quickly

established, even though equilibrium burnup con- ditions did not exist in the reactor. Fuel manage- ment was based on a 3-D diffusion calculation of burnup with one mesh point per fuel bundle and it was found that close to optimum burnup could be obtained by the fairly simple principle of removing the most highly burnt-up fuel bundle when more reactivity was needed. Some minor modification of this scheme was necessary to ensure uniform fuelling over the reactor face and in the two refuelling directions. Thus, while fuel management was a continuing operational requirement, planning was fairly simple and did not require high precision calculations since only a small fraction of a channel's fuel was changed at one time and reactor and channel instrumentation could detect any anomalous trend in behaviour which could be compensated in subsequent fuel changes.

This simple fuel management scheme was com- pletely successful for the NPD reactor which required only one new fuel bundle per day, but experience with NPD indicated that single bundle shifts for full size power reactors would require excessive fuelling machine operation.

F U E L M A N A G E M E N T I N D O U G L A S P O I N T

Two-bundle fuel shifts--i.e, inserting two bundles at each refuelling operation--were originally planned for the Douglas Point 200 MWe reactor to reduce the number of channel visits. (The fuelling machines are capable of refuelling up to a full channel [12 bundles] at a time.) However, studies showed that four-bundle shifts gave negligible additional burnup penalty so these were adopted to further reduce fuelling machine usage. Some fuel failures occurred and were shown to be caused by moving a partly irradiated fuel bundle from a low flux

359

360 C . H . M~Lt~

Fig. 1. CANDU refuelling concept.

position to a higher flux position (Robertson, 1972). The appearance of this effect for the first time in Douglas Point was due partly to the higher power rating of the fuel (compared to NPD) and partly to the larger step increases in neutron flux seen by the fuel as a result of multi-bundle shifts. This is illustrated in Fig. 2 where it is seen that with a four- bundle shift, bundles 1-4 in a channel are all moved to positions of higher flux, with the effect being particularly marked for bundle 1. All other bundles are moved to positions of lower flux, or out of the

reactor. The allowable flux increase as a function of integrated bundle irradiation was established empirically and, while fuel designers devised improved fuel (Bain et al., 1975), the refuelling scheme was changed to 8-bundle shifts. This mzt the new flux increase criterion, and reduced attain- able fuel burnup only about 5 ~o.

FUEL MANAGEMENT IN PICKERING

Initial experience

Unit 1 of Pickering (540 MWe) was started up in 1971 before this fuel failure mechanism had been fully identified. In the initial planning three factors were used to determine the fuelling economics:

Fuel makeup costs, that is, the cost of the fuel bundles inserted into the reactor per unit energy output,

Fuelling machine operating and maintenance cost, and

Fuel defect cost, consisting of the lost burnup, the added fuelling machine operating and mainte- nance cost, and the cost of radiation dosage to maintenance staff. It was assumed that the sole cause of fuel failure due to refuelling would be the large increase in power of highly irradiated bundles after fuel shifting. On this basis 8-bundle shifting appeared the most economical and was therefore adopted as the planned fuelling pattern. The burnup penalty associated with this pattern was 10 MWh/kgU, yielding a net expected burnup of 152 MWh/kgU for Unit 1.

NEUTRON FLUX (RELATIVE)

FINAL STATUS

(A) AFTER 4-BUNDLE SHIFT I I I I 2 3 4 I '

(B) AFTER 8.BUNDLE SHIFT I I i l I 2 3 4 5

LEGEND

I IFRESH FUEL ~ O N C E . S H I F T E D FUEL 2 2'

INITIAL POSIT ION IN CHANNEL

! Z , 3 , 4 , 5 , 6 , 7 , 8 I 9 , 1 0 , I I , 1 2 ,

FUEL BUNDLE POSIT ION IN CHANNEL

2'

I I 6

3' 4 ' 1" 2" 3" A"

I 7 8 1' 2 ' 3 ' . 4'

TWICE.SHIFTED FUEL 2"

Fig. 2. CANDU multi-bundle fuel shifts.

Fuel management in

In addition to the bundle power increases due to refuelling, a sequencing error in the withdrawal program of the neutron-absorbing "adjuster rods" during reactor start-up caused additional local increases in some regions and a number of fuel failures occurred. Computer studies using the S O R G H U M 3-D, 2-neutron-group reactor kinetics code (Akhtar, 1972) showed that during some startups of Unit 1, some bundles had experienced neutron flux increases of up to 47 ~o above their normal operating level, and in some cases this resulted in bundle operating powers as much as 22 ~o above the design rating of the fuel. Even after correct sequencing of absorber rod removal had been established, the calculations showed that some bundles had experienced powers 36 ~o above normal. The fuel defects were attributed to these large power increases.

To reduce the number of defects due to adjuster rod operation, two steps were taken. First, adjuster rod removal sequence and allowable reactor power levels during trip recovery were revised using three ground rules (Jones and Turcotte, 1974):

1. Maximum bundle power would be limited to 15 ~ above normal.

2. As reactor power increased, adjusterrodswould be inserted as close as possible to the location of the maximum power increase.

3. The reactor power would be held constant or increased during the recovery period, but never reduced.

Secondly, the fuel shifting procedure was modified so that the bundles from positions 1 and 2 which originally spent 18 rain in the high flux region of the channel during the change now spent only 5 rain there.

It was also necessary to reorganize the fuel management strategy. Complex empirical flux

CANDU reactors 361

increase criteria were established on the basis of experience in Douglas Point and Pickering and a development program at Chalk River. These are illustrated in Fig. 3.

The fuel element defect threshold after a power increase was set as a decreasing linear function of integrated irradiation. In addition, the step increase in power rating of the outer fuel elements was to be kept below the appropriate member of the family of curves shown on the AP graph in Fig. 3. During the multi-bundle shifts some of the fuel bundles must toler~ite the high neutron flux at the centre of the reactor, but it was found that the fuel can tolerate greater step power increases for a short time than it can on a permanent basis--hence the family of curves for different periods at high power. If both these criteria are violated, experience has shown that fuel failure is likely. The arrows on the figure show examples of fuel manoeuvres that would result in defects and those that are allowable.

The economics of 8, 10 and 12-bundle shifting in the high power region were re-examined (Pasanen, 1973) in light of the increased cost associated with fuel defects, and it was concluded that 10-bundle shifting was the most economic so this was instituted. This decreased the final power level of the bundles from positions 1 and 2, which now moved to positions 11 and 12.

This ability to change the fuelling pattern to meet an unforeseen fault situation allowed, in this instance, fairly continuous operation of the reactor with minimum down time, minimum fuel wastage, and only a few weeks of reactor derating to prevent excess power operation of the freshly fuelled channels.

This flexibility of the fuel management was again demonstrated in the fall of 1974 in the Pickering Unit 1 reactor. The fuelling tnachines became

==~ 50

" : u ~ Z z = ao

, ~ 2o

..=.< lo

o2.<

FUEL COULD DEFECT WHEN FUEL SHIFT VECTOR CROSSES BOTH p &Ap LIMIT LINES

l l I

HIGH PROBABILITY OF DEFECT

DEFECTS +UNLIKELY

5b ' I00 150 J 2OO

w, OUTER ELEMENT BURNUP IMWh/kg.U)

-+oi 40

~= 30 +°[ = z 20

go °o s'o 100 150 200

w, OUTER ELEMENT BURNUP (MWh/kg-U)

Fig. 3. Fuel shift criteria for Picketing.

362

Table 1. Picketing operation with adjuster rods with- drawn

Number of adjuster rods withdrawn

Maximum allowable reactor power level

(~o full power)

1-2 97 4--6 88 8-14 80

16 73

inoperable for nearly three months and reactivity had to be maintained by progressive removal of adjuster rods. Early experience had shown that some derating of the reactor is required to operate with adjuster rods removed. Table 1 shows this in tabular form. The loss of productivity during this period is illustrated graphically in Fig. 4.

Current practice The fuelling pattern used at Pickering is a flexible

one (Brown, 1974). Fuelling is planned a few days in advance and a specific list of channels to be fuelled is issued to the operating crews who perform the fuelling operation.

Channels are selected for fuelling on the following basis. The status of the reactor core is determined at the end of each month by the 3D, 2-group diffusion code SORO which simulates the conditions throughout the reactor core on the basis of the average value of the zone control absorber levels,

C. H. Mn.LAR

adjuster rod positions and fuel shifting records sampled at intervals of approximately 250 GWh(th) during the month (about once a week). Flux distributions, bundle and channel powers, bundle irradiations and burnups are calculated and a listing produced of channel burnups in descending order for each fuelling region. This channel burnup list is the prime priority list for fuelling the reactor.

Several other modifying factors, however, are also considered:

High reactivity gain per channel. Sufficient re- activity is normally maintained by fuelling at a rate to just compensate for the reactivity loss due to burnup. When this rate cannot be maintained, it is necessary to fuel channels that will produce the highest reactivity gain per channel visited.

High temperature regions. Fuelling in regions of high channel outlet temperature is deferred if it is likely to produce a further increase in temperature and a subsequent reactor derating.

Symmetry. The reactor is fuelled symmetrically about the cylindrical core axis with equal number of channels fuelled per control zone. Axial symmetry is maintained by fuelling alternately east and west fuelled channels.

SUMMARY OF FUELLING EXPERIENCE IN PICKERING

The success of the fuel management program and some of the lessons learned from it can be illustrated

I - -

REACTOR POWER

(%)

0 20 40 60 80 100 120

TIME (DAYS)

Fig. 4. Pickering productivity loss during period of fuelling machine unavailability.

Fuel management in

by examining operating data from the Pickering station.

Maximum bundle power history

Figure 5 shows the history of the maximum bundle powers as a function of total reactor output for all four Pickering units. The shaded band is a 4-10 % variation about the nominal design maximum of 640 kW per bundle, the upper end of the band being the maximum bundle power limit of 705 kW. The objective of the fuel scheduling scheme is to maintain the maximum bundle power close to the 640 kW value and the data show that, in general, this has been achieved. There have been four occasions where the maximum bundle power has exceeded the 705 kW limit.

The first excursion of Unit 1 at --~1.5 TWh(th) was caused by a transient following withdrawal of a shutoff rod which had been stuck in the core.

At approximately 8 TWh(th) fuelling began in Unit 1 and the maximum bundle power started to rise. The increase had levelled off below the 705 kW level by approximately 10.5 TWh(th). It was during this time that the defects due to out-of-sequence adjuster rods were being removed. Coincidentally with this, there were problems with the running of the FASP code (the forerunner of SORO) and the routine monthly outputs were very late. As a result, fuelling continued and even though depleted bundles had been used to reduce bundle powers in the area with defects, the 705 kW limit was exceeded and discovered after the fact.

CANDU reactors 363

Up to this time the only temperature limitation in effect was a bulk one for the whole core (to limit overall reactor power) and an alarm on the fully instrumented channels to avoid significant boiling in the channels. The preceding events indicated that operating limits should be derived for temperature rise across each channel (AT) to ensure that the maximum rated bundle in the channel does not exceed 705 kW. These AT limits are shown in Fig. 6. They were derived by examining the past history of flux shapes in each channel and using these distributions, with the maximum bundle power normalized to 705 kW, and the allowable AT was calculated based upon design flow in the channel. The smallest allowable AT found for any channel in each of the zones shown in Fig. 6 was set as the upper limit of ATfor any channel in the zone. The reactor is normally operated so that these AT's are not exceeded.

The third excursion in Unit 1 occurred in the fall of 1974. Unit 1 was inadvertently overfuelled in some regions, and this, together with deviations of axial flux distributions (due to an attempted return to 8-bundle shifts with improved fuel) and variations of channel flows led to some bundles being over- rated despite the fact that AT limits had not been violated. Refined AT limits and operational prac- tices were instituted to minimize the probability of this situation re-occurring.

The 705 kW limit was exceeded in Unit 2 at ,---17.5 TWh(th) and was caused by inadvertent clustering of newly fuelled channels during a period

0

Q z m

<

i!i!

V UNIT 4 i

o lo 15 2'o 2's 3'o 3'5 TOTAl HEAT OUTPUT (TWhith))

Fig. 5. Picketing bundle power history.

364 C. H. MRa.AR

e ummmmmmmmmm emummmmmmmmmm ue = eemmmmmmmmmm m eemnmmmnnme e

m m emmmmmmmme

m 50°C ~ 49oC IT1"i'TI 47oC

Fig. 6. AT limits for Picke, dng Unit 1.

of rapid refuelling to correct for a known reactivity deficiency. Deratings were introduced because of the AT limits being approached but were not enough to prevent some overrating.

That this evolving fuel management system has been successful is shown on Table 2 where it may be seen that since the initial problems in Unit I, fuel failures have been minimal.

Maximum channel power history

Figure 7 shows the maximum channel powers for all four units. As seen from the data, channel power changes have, with a few exceptions, been main- tained within + 10% of the 5.5 MW time-averaged design value.

Fuel consumed

Figure 8 shows some of the features of fuel usage for Units 1 and 2. The band in the lower portion of

the figure shows the predicted fuel usage for the 150- 170 MWh/kgU range of burnups indicated. The upper portion of the figure shows the actual average monthly discharge burnup and average in-core burnups for the respective units.

The data for Units 1 and 2 show that the initial prediction of 152 MWh/kgU was low and that a prediction made approximately one year ago of 160 MWh/kgU, based upon operating experience to that time, is still low. If we assume that "equi- librium" was reached at about 20 TWh, then the current best estimate for burnup is 172 MWh]kgU.

OTHER CONSIDERATIONS IN CANDU FUEL MANAGEMENT

Booster rods vs adjuster rods

One facet of C A N D U fuel management is associated with the method of overcoming the

Table 2. Pickering fuel performance to end of 1974

Number of fuel bundles

Irradiated Discharged Defective ~ Defective*

Unit I Before Nov. 1, 1972 6,938 2,258 91 1.31 After Nov. 1, 1972 12,754 8,074 3 0.02

Total-Unit 1 15,012 10,332 94 0.63 Unit 2 13,646 8,966 1 0.01 Unit 3 9,818 5,138 6 0.06 Unit 4 9,296 4,616 Nil Nil Total-Pickering 47,772 29,052 101 0.21

* Per cent defective bundles = Total discharged defective bundles × 100~o Total irradiated bundles

Fuel management in CANDU reactors 365

MW 7

U

UNIT1

I 10 20 30 0

Fig. 7. Pickering channel

Xe-135 absorptivity transient following a reactor shutdown. In a natural uranium reactor it is not feasible to incorporate sufficient excess reactivity to overcome the full Xe-135 transient, but enough can be provided to allow start-up within about a half- hour of shutdown. In the NPD and Douglas Point reactors a set of "booster" rods of aluminum/93 ~o- enriched uranium can be inserted when additional reactivity is required. However, in Pickering, cobalt "adjuster" rods are inserted normal to the fuel over the central region of the core for removal when extra reactivity is required. During normal operation these represent a reactivity loss of slightly less than 2~o, but this is not totally reflected in reduced

f AVG. IN CORE BURNUP

I I I I I , I

200

15o

loo

so z

0

150 - - - P 1 / / / .

"' S / 15o M w h / k g T / if,

~ " , ) , , , I ' I 0 0 ~a j , /~ l r0 Mwh/kgu

FUE 7 50

0 0 5 lO 15 20 25 30

TOTAl. HEAT OUTPUT (TWh[th))

Fig. 8. Fuel consumption and burnup for Pickexing units 1 and 2.

UNIT2 UNIT3 UNIT4

75=/0 FP I r I I 2 1 0

10 20 3 0 0 10 3 0 0 )0 20 30

TOTAL HEAT OUTPUT (TWh(th))

power history and bumup.

burnup. With the booster rod system radial flux flattening is produced by irradiating the fuel in the central channels to higher burnups resulting in a larger fission product load in this region. Adjuster rods replace some of this fission product load to give some radial flux flattening and also some axial flattening. This produces a more uniform burnup distribution in the reactor core with a consequent reduction in channel-to-channel power "ripple" from 15Yo to about 5% (Pasanen, 1973).: This uniformity makes for higher average irradiations. The net result is that a reactor using adjuster rods suffers about a 10~, penalty in burnup.

Initial transient suppression

Suppressing the excess reactivity of the full charge of fresh fuel at initial start-up is not a major problem. In the NPD reactor some depleted fuel bundles were spaced throughout the core for this purpose but an accidental downgrading of the D20 early in the reactor operation forced their early removal. For the later reactors, designers have favoured boron poison in the moderator to suppress excess reactivity at any time in the reactor's life, and the necessary boron injection and ion-exchange removal systems are a permanent part of the reactor equipment. In reactors using booster rods, howevca', the uniform suppression of reactivity by moderator poisoning provides inadequate flux flattening during the initial reactivity transient to allow full power operation without overpower operation of some channels. In this case, some of the necessary absorption is provided by judiciously-located bundles of depleted fuel. Once past the initial transient the boron poisoning mechanism provides a useful fuel management tool, allowing, for example, excess reactivity to be stored in the reactor by premature refuelling ff the fuelling machine must be removed for maintenance.

366 C. H. M n ~

Vertical channel CAND U-BL W

The characteristics of the 250 MWe CANDU- BLW reactor at Gentilly, Quebec, present new fuel management problems. First, the lattice character- istics differ markedly above and below the boiling boundary in the vertical fuel channels, and the coolant flow is upward in all channels. Secondly, on-power fuelling is by a single fuelling machine underneath the reactor which completely withdraws and replaces a full channel of fuel. Some of the bundles withdrawn are reassembled with fresh ones to form a new fuel string for use in a different channel. The 10-bundle fuel string is mounted on a central tie rod and is manipulated under water in a horizontal position. This allows bundles to be shifted to any site along the recycled fuel string to cope with the vertical flux asymmetry caused by the boiling coolant. Because of the extended shutdown of the Gentilly reactor due to a D20 shortage, little experience in fuel management has been obtained.

Enriched fuel cycles

Advanced C A N D U systems will need more sophisticated fuel management. In a recent design study for a CANDU-BLW reactor with plutonium enrichment, fuel management was complicated by the fact that the fuel bundle power decreased by 30 Yo over its lifetime while in a comparable natural uranium C A N D U the bundle power remains within 4-5 % of its mean value. Since reactivity varies in a similar manner, one- or two-bundle fuel shifts to maintain constant reactivity would be preferable but would require excessive fuelling machine

operation. The solution proposed was to discharge the whole fuel string (10 bundles) at each refuelling-- which reduced channel visits tenfold--but to add axial reflectors to give additional axial flux flattening and help increase average burnup of the fuel string. However, the large local-increase in reactivity caused by the insertion of a whole string of fresh fuel required careful sequencing of channels refuelled to minimize local power peaking.

Studies on the use of thorium in C A N D U reactors show that the reactivity "bold-up" in Pa-233 makes the operating power of a fuel bundle a sensi- tive- function of its past irradiation history and current flux level. This adds yet another dimension to the fuel management problem.

Acknowledgments--The author wishes to acknowledge the major contribution to this paper of Dr. R. A. Brown, Central Nuclear Services division of Ontario Hydro. The details of Pickering fuel management principles and operational experience together with most of the figures and tables are from a colloquium presented by Dr. Brown at Chalk River February 17, 1975. The assistance of staff members of AECL Power Projects division is also gratefully acknowledged.

REFERENCES

Akhtar P. (1972) Private communication. Bain A. S., Page R. D. and Fanjoy G. R. (1975) 1st

European Nuc. Conf. Barss W. M. (1962) AECL-1528. Brown R. A. (1974) CNA-74-TW-3. Jones J. E. and Turcotte G. E. (1974) Trans. Am. NucL

Soc. 18, 289. Pasanen A. (1973) IAEA-SM-178/13. Robertson J. A. L. (1972) EngngJ. (Nov./Dec. 1972).

pp. 9-13