New Method For Reactor Core Loading Pattern Optimization

In China, with the rapid development of nuclear power, the corresponding task ofreactor core reloading design will be more and more. Reloading design is based on thecore loading pattern optimization. In order to develop practical engineering optimizationmethod and program, this thesis focused on this problem, and the research content isshown as below:Firstly, this thesis proposed a new heuristics method-interval bound algorithm(IBA), which is specifically used for loading pattern optimization. Its calculationprocess was described in detail. IBA directly uses the reactivity of fuel assemblys andtheir four quadrants and the worth of burnable poison to carry on evolutionarycalculation. It is more intelligent, simple realization and inherent parallelism. It canoptimize fuel assembly location, burnable poison placement and used fuel assemblyorientation in a coupled way. The numerical experiment in several types of problemsproved the optimization ability of IBA.Secondly, IBA was theoretically analysed and improved. The definition of controlvariables is indicated necessity and superiority. It is reasonable that intervals are used asthe model to characterizing learning rule. Based on contractive mapping principle, theglobal convergence of IBA was proved. Adding the elitism strategy for IBA can balancethe selection pressure and population diversity better, so both the convergence speedand optimization solution quality have been improved. To deal with multi-modalloading pattern optimization problems, we also added multi-interval-model for IBA, andthen the algorithm can solve multi-modal optimization problem without increasingcomputation.Thirdly, genetic algorithm (GA) and interval bound algorithm were compared inmono-modal and multi-modal optimization problem respectively. In mono-modaloptimization problems, IBA is more capable to generate the population diversity, takesinto account the linkage between variables, and uses the heuristic information such asthe reactivity of fuel assemblies, so its performance is better than GA. In multi-modaloptimization problems, in addition to the previous three reasons, genetic drift makesthat GA is easy to only converge to a suboptimal modal, so the performance of GA is worse than IBA.Finally, to deal with multi-objective constrained loading pattern optimizationproblem, this thesis proposed dynamic discontinuous weight factors (DDWF). Thistechnique can co-ordinate the constraints and multiple objectives. The weight factorschange with the evolutionary process to meet different selection pressures for infeasiblesolutions at different evolutionary stages. The weight factors are discontinuous in thelimit values can reduce the sensitivity of fitness function to the factors. Through theconversion formula of various constraints and objectives, engineers can express theirpreferences on multiple objectives, and the infeasible solutions can also be effectivelypunished.Integrating the above research results, we developed a pressurized water reactornuclear power plant core loading pattern optimization program IBALPO. Benefit fromthe good performance of IBA and DDWF, IBALPO has strong practicability, and hasbeen used to help engineers to optimize the loading patterns in NPIC.