Present status of reactor physics in the United States and Japan-III. 1. Recent Activities of Loading Pattern Optimization Research in Japan

Because of recent enhancements of optimization algorithms and great improvements in computer hardware, loading pattern (LP) optimization methods are being used as practical design tools both in the pressurized water reactor (PWR) and boiling water reactor (BWR) industries. LP optimization methods are mainly used for the following in-core fuel management activities in Japan: 1. minimization of fuel cycle costs; 2. evaluation of various in-core fuel management scenarios; 3. estimation of the number of feed assemblies needed during several successive cycles for fuel ordering; 4. evaluation of fuel bids. Although engineers can perform these analyses, the major motivations to utilize LP optimization methods are the reduction of manpower and the establishment of engineer-independent LP quality. These are important in today's in-core fuel management tasks. In the following sections, activities related to LP optimization research in Japan are briefly described. The major activity of PWR LP optimization research in Japan is development of the INSIGHT system. The INSIGHT system is an integrated scoping analysis tool for PWRs developed by Nuclear Fuel Industries (NFI). The INSIGHT system is a graphical user interface (GUI)-based interactive design tool that includes LP optimization, automated multicycle analysis, an interactive LP design, core follow, an integrated database, and some auxiliary functions. The INSIGHT system was mainly written in the C++ language and consists of ∼400 000 lines of source code. The GALLOP code is the LP optimization module of the INSIGHT system. An automated multicycle analysis is performed by the MCA code in INSIGHT. The MCA code performs a fuel and burnable poison (BP) inventory search by automatically invoking the GALLOP code, which makes LPs. The MCA code can deal with various constraints that have appeared in practical in-core fuel management, e.g., limitations of fuel/BP stock, forced fuel loading/discharge, limitations of core safety parameters in each cycle, and reinsertion of fuel/BP. Although input data for both of the GALLOP and MCA codes are not short (typically more than 100 to 200 lines), these input data are automatically generated by the INSIGHT system through its GUI operation. The INSIGHT system is independent from the core calculation code. Currently, the SHARP code, which is NFI's in-house core calculation code, and SIMULATE3 can be used as a core simulator. BWR optimization is much more difficult since its design space is significantly larger than that of a typical PWR, and three-dimensional core calculations are necessary because of strong thermal-hydraulic feedback effects. However, Toden Software has recently developed FINELOAD3, which performs BWR LP optimizations. The FINELOAD3 code utilizes the modified tabu search method coupled with fast core calculation capability using the simple linear perturbation (SLP) method and detailed three-dimensional core calculations. The SLP method estimates the in-core power distribution based on the superposition of the assembly power change obtained from the power/multiplication sensitivity library generated in a reference equilibrium core. In FINELOAD3, the optimization search starts from a user-supplied initial LP. To perform efficient optimization, assembly binary shuffling was carried out based on a priority table that is evaluated by contribution to the violation of safety parameters. This method is currently used in actual reload core design tasks in TSI. SIMULATE3 is used as a core simulator in the FINELOAD3 code. Toden Software is also trying another optimization approach using genetic algorithms in which the genetic operation was modified to improve the search efficiency. This approach does not require an initial LP for the starting point, and its search capability is robust. However, numerous core calculations are required to obtain a converged solution. Computation time is generally critical in practical applications of the LP optimizations. Therefore, the reduction of computation time is strongly desired from a practical point of view. This can be achieved by a speedup in core simulators, parallel computation, and improvements in optimization algorithms. In a PWR reload core design, experienced engineers can reach their final LPs through dozens or hundreds of trial LPs if fuel inventory is fixed. On the contrary, common stochastic optimization algorithms require the evaluation of thousands of LPs to obtain converged solutions. If an 'engineer-like optimization method' that utilizes some kind of reasoning is developed, it will greatly contribute to the reduction of computation time. This may be achieved by case-based reasoning or knowledge-based reasoning algorithms. The accuracy of a core simulator is also important. If the prediction accuracy of a core simulator is poor, LP optimization will fail to obtain near-optimum solutions. Therefore, continuous efforts to develop more efficient core calculation methods are necessary. Current LP optimization codes provide near-optimum solutions with minimum user input; they act like a 'black-box' pattern generator. However, from our experience, the black-box optimization code has a potential risk: Its use can make in-core engineers less alert and may have a negative impact on overall in-core fuel management tasks. Therefore, adequate interaction between engineers and LP optimization codes would be desired in future in-core fuel management tasks. (authors)