Application of differential evolution algorithm for fuel loading optimization of the DNRR research reactor

Abstract

A discrete differential evolution (DE) algorithm has been developed and applied to the problem of fuel loading optimization of the Dalat Nuclear Research Reactor (DNRR). The DNRR is a 500 kW pool-type research reactor loaded with Russian VVR-M2 fuel type. The discrete DE algorithm is applied to the fuel loading pattern optimization of the DNRR core consisting of 100 highly enriched uranium fuel bundles with different burnup levels. A fitness function was chosen for maximizing the effective multiplication factor, keff, and minimizing the power peaking factor (PPF). The DE search was performed with ten independent runs through 500 generations with the population size of 30. The results show that the fitness function approaches a maximum value after about 450 generations, while the maximum keff reaches a stable value after 250 generations, and the PPF achieves minimum value after 150 generations. Three best loading patterns (LPs) have been selected from the DE search. Compared to a reference LP, i.e. a working core configuration of the DNRR, (keff=1.06040 and PPF=1.374) the best LPs correspond to the keff values greater by about 470–495 pcm, while the PPF values are smaller by about 3.7–4.0%. Comparing with the genetic algorithm in a previous work, the DE is more advantageous in exploring search space and approaching a global optimization.