Adaptable Long-Term Optimization of Dry Cask Storage Loading Patterns

Abstract

To address the evolving needs of dry storage, this research developed an optimization methodology to identify loading configurations to minimize the number of casks, their heat load, and the time when they meet transportation requirements. The motivation was to investigate strategies that balance and reduce risk over the lifetime of a site's reactor(s).

The dry cask loading problem was formulated as an adaptable dynamic bin packing problem, accommodating different site and cask limits in broadly-defined constraints. A new method was developed to address its complexities, named the GRASP-enabled adaptive multiobjective memetic algorithm with partial clustering (GAMMA-PC). This method embeds greedy randomized adaptive search procedures in a multiobjective evolutionary algorithm with local search techniques and partial decomposition of the objective space during crossover.

GAMMA-PC was demonstrated through integration with the unified database from the Used Fuel Systems group at Oak Ridge National Laboratory to optimize simulated loading configurations for Vermont Yankee, Comanche Peak, and Zion Nuclear Power Stations. Its performance was evaluated through comparisons to test solutions and to the real Zion loading configuration. GAMMA-PC produced diverse solutions that dominated the testing sets. The improvement was concentrated in the average heat load, and the third objective function was shown to be sensitive to individual assembly characteristics.

The results suggested the usefulness of GAMMA-PC for utilities considering long-term goals. They showed that more diverse cask loadings and strategic placements of empty positions can be used to reduce initial heat loads. Moving to a higher capacity cask increases loading flexibility but can result in transportation delays. Long-term planning enables a more thorough consideration of the trade-offs involved in any decision.

This research contributes one of the first in-depth studies of the dry cask loading problem. It expands the current treatment of assembly selection over longer timeframes and meets user-defined requirements. It is also one of the first tri-objective dynamic bin packing problems, and the first to pack items with time-dependent characteristics. Future work should focus on refining the objectives and incorporating uncertainty. With its adaptable structure, GAMMA-PC is a promising new metaheuristic for this task and for dynamic bin packing problems in general.