Intelligent stochastic optimization routine for in-core fuel cycle design
暂无分享,去创建一个
Any reactor fuel management strategy must specify the fuel design, batch sizes, loading configurations, and operational procedures for each cycle. To permit detailed design studies, the complex core characteristics must necessarily be computer modeled. Thus, the identification of an optimal fuel cycle design represents an optimization problem with a nonlinear objective function (OF), nonlinear safety constraints, many control variables, and no direct derivative information. Most available library routines cannot tackle such problems; this paper introduces an intelligent stochastic optimization routine that can. There has been considerable interest recently in the application of stochastic methods to difficult optimization problems, based on the statistical mechanics algorithms originally attributed to Metropolis. Previous work showed that, in optimizing the performance of a British advanced gas-cooled reactor fuel stringer, a rudimentary version of the Metropolis algorithm performed as efficiently as the only suitable routine in the Numerical Algorithms Group library. Since then the performance of the Metropolis algorithm has been considerably enhanced by the introduction of self-tuning capabilities by which the routine adjusts its control parameters and search pattern as it progresses. Both features can be viewed as examples of artificial intelligence, in which the routine uses the accumulation of data, or experience, to guidemore » its future actions.« less