Nuclear Science and Engineering / Volume 193 / Number 5 / May 2019 / Pages 523-536
Technical Paper / dx.doi.org/10.1080/00295639.2018.1550970
The constraint annealing method is presented and demonstrated for the solution of large-scale, multiconstrained problems in light water reactor fuel cycle optimization. Constraint annealing is a penalty-free method that eliminates the need for traditional constraint weighting factors by treating each objective function and constraint as separate and concurrently solved minimization problems within a global optimization search framework. The current application seeks to demonstrate the effectiveness of constraint annealing for a complex core loading pattern design problem containing multiple objective functions and constraints without the need for additional ad hoc control parameters. Two problems of varying degrees of complexity are analyzed. The first problem is defined by a single objective function based on maximizing cycle energy with two constraints based on power peaking and peak rod exposure. The second problem expands upon the first by adding an additional objective function for vessel fluence and four additional constraints based on controlled power peaking, steaming rate, moderator temperature coefficient, and alternate source term. Results demonstrate that constraint annealing inherently addresses issues of scaling associated with different objective function and constraint formulations as well as the impact on cycle energy.