Parallel Adaptive Evolutionary Algorithms for Pressurized Water Reactor Reload Pattern Optimizations

Abstract

The combination of traditional evolution strategies and heuristics from expert knowledge leads to the RELOPAT optimization program. In combination with reactor simulation codes—in this investigation the nodal reactor code PRISM of Siemens/KWU— a powerful program system for the design of a numerically optimized pressurized water reactor (PWR) loading pattern was designed. Furthermore, the technic of parallel computing was introduced successfully. Simple parallel algorithmic structures on the level of optimization algorithms, combined with a low amount of communication between processors, allow workstation clusters to be used efficiently. Highly promising results were obtained by comparing recalculations of different known loading patterns for several PWRs of different sizes and varying constraints with solutions based on human expertise. The economic potential of the improvements now leads to a program that meets industrial requirements.