Optimization of fuel reload for a BWR using neural networks and genetic algorithms

In this work we use a backpropagation neural network (NN) and a genetic algorithm (GA) to optimize the nuclear fuel reload in a BWR reactor. The NN was trained with a fuel reload set evaluated previously with a reactor simulator code [1,2], to estimate the k(eff) and thermal limits values and other parameters. The GA generates fuel reload patterns searching the optimal one. We use two criteria to qualify a fuel reload: the requirements or constraints at the Beginning of the cycle (BOC) and at the End of the cycle (EOC). From a theoretical point of view the fuel reloads satisfies the operational constraints for a BWR reactor similar to Laguna Verde Nuclear Power Plant in Mexico.