CANDU refuelling optimization by successive two-step mathematical programming - I: MILP models for refuelling channel selection

A CANDU refuelling optimization method based on successive two-step mathematical programming (MP) is developed. The first step is for the selection of weekly refuelling candidates, while the second step is for the determination of the detailed weekly refuelling scheme with the prescribed candidates. To search for the overall optimal solution with the aftereffects of refuellings being considered, a zero- dimensional core physics model and a three-dimensional core physics model are directly incorporated into their corresponding MP model. The formulated large-scale MP problems are solved by use of the powerful generic MP problem solver CPLEX® (commercial product of IBM ILOG). The potential for engineering application of the method is demonstrated by its on-site test application to Qinshan CANDU 6 reactor. Refuelling simulation against the operation history of 392 full power days demonstrates that in addition to the good regional power and core reactivity ripple control, higher average fuel discharge burnup of 176.54 MWh/kgU is achieved. No violation of reactor operating limits is found. Regarding the computational cost of the method, though for most tested cases the performance of the method satisfies the stringent on-site time requirement, some exceptional cases are observed. However this occasional speed problem is within a factor of two or three from the targeted requirement, which is realistically optimistic in the near future considering today's rapid advancement in hardware and software.