Optimal Fuel Management of CANDU Reactors at Approach to Refuelling Equilibrium

Abstract

Initially the CANDU Pressurised Heavy Water Reactor (PHWR) was designed to use natural uranium dioxide (U09 as its fuel, but its versatility was soon evidenced as early studies demonstrated that the reactor could accommodate several other types of nuclear fuels without the need for major re-design. These alternate fuels include low enriched uranium (LEU), thorium-containing oxide mixtures (MOX), plutonium-based MOX, and Pressurised Water Reactor (PWR) spent fuel recycled in CANDU reactors. CANDU reactor refuelling is carried out on-power in a quasi-continuous fashion by having a refuelling machine inserting a string of fresh fuel bundles into one end of a he1 channel and another similar machine unloading an equal amount of spent fuel bundles. The quasi-continuous nature of the refuelling creates a so-called refuelling equilibrium state, which is reached about one year after initial reactor start-up and characterized by a constant rate of refuelling and a constant discharge burnup. The present work rather focuses on the "Approach to Refuelling Equilibrium" period which immediately follows the initial commissioning of the reactor. This period is indeed a sequence of two phases: the first one characterized by the absence of refuelling since the reactor has initially a high excess reactivity compensated by burnable poison present in its moderator, and the second one when refuelling is done at a higher rate which tapers out to the constant value of the equilibrium The in-core fuel management problem for this period is treated as an optimization problem in which the objective function is the refuelling frequency to be minimized by adjusting the following decision variables: the channel to be refuelled next, the time of the refuelling and the number of fresh fuel bundles to be inserted in the channel. The optimization problem also includes the following constraints: maximum channel power, maximum bundle power, maximum discharge burnup, minimum time span between successive visits by the refuelling machines, and, in some cases, a minimum channel separation for consecutive refuellings. In this study, a simulation program was developed to investigate the optimum approach to refuelling equilibrium in a CANDU-6 reactor. Finite difference methods are used to solve the diffusion equations in two energy groups, using a 80 x 80 x 26 mesh point grid yielding 18 mesh points per fuel bundle The lattice parameters are determined from interpolating neutron flux and fluence within parameter tables calculated using the transport code WIMS-AECL coupled with the ENDFfB-V data library. Refuelling is simulated by setting the fluence to zero for mesh points representing fresh fuel bundles. Since at this stage it is desired to investigate the effects of the refuelling optimization itself, the model does not include reactivity control devices for the moment. The model simulates individual refuellings and follows closely the evolution of the core throughout the whole "Approach to Refuelling Equilibrium" period. Channel selection is optimized by first setting up a modest group of channels as candidates for the next refuelling, based on accumulated fluence and on exclusion criteria when a channel separation constraint is active. Perturbation Theory is used to simulate in turn the refuelling of these channels and to calculate for each of them, among other parameters, the cycle length (i.e. the time between this refuelling and the next). The channel for which refuelling yields the largest cycle length is then selected and its refuelling "implemented" in the full core model. This process is repeated until the whole "Approach to Refuelling Equilibrium" period is simulated and optimized. In this study, two methods of refuelling were investigated: a so-called "static" method in which the refuelling mode (i.e. number of fresh fuel bundles inserted) and the minimum channel separation constraint were initially selected and unchanged throughout the simulation, and a "dynamic" refuelling method in which the two parameters above were part of the decision variables. With this last method, "deshifting" was observed in some of the simulations as it permitted to resolve maximum power density constraints. Four different fuels were investigated in this work: natural U02, the lower fissile content plutonium MOX option investigated at AECL for nuclear weapon's material disposal in a BRUCE A CANDU reactor, "Self-Sufficient-Equilibrium Thorium Cycle" (SSET) fuel as an ultimate case representative of thorium-containing fuels, and DUPIC ("Direct Use of Spent PWR Fuel in CANDU") fuel. Both the present 37-rod bundle design and the proposed CANFLEX bundle design are part of this study. The results include the time to reach refuelling equilibrium from initial start-up of the reactor, the average discharge burnup, the average refuelling frequency and the average channel and bundle powers relative to natural UOi. The model was initially tested and the average discharge burnup for natural U02 came within 2% of the industry accepted 199 MWh/kgHE. For this type of fuel, the optimization exercise predicted the savings of 43 bundles per full power year. The optimization yielded for the advanced fuels the following values for the average discharge burnup at refuelling equilibrium: PuMOX: 180-186 MWh/kgHE, SSET: 492-503 MWh/kgHE, and DUPIC: 635-638 MWh/kgHE. The calculations also evidenced some problem areas such as high power densities for fuels such as the DUPIC. In this case in particular, frequently, the optimization had to be set aside in order to refuel channels in which bundles had reached or exceeded the maximum allowable burnup. These problems would be resolved once the model includes a representation of the control elements. Perturbation Theory has proven itself to be an accurate and valuable optimization tool in predicting the time between successive channel refuelling operations.