Parametric Optimization of TRU Destruction Rates in HTR Cores Using Hybrid Metaheuristic Algorithm and Regression Methods
Abstract
Nuclear waste is one of the significant problems for the sustainable use of nuclear energy. The Deep-Burn using High-Temperature Reactor (HTR) is one strategy to minimize nuclear waste. The nuclear waste or TRU incineration problem in HTR is one of the complicated optimization problems because of the broad design parameter space and constraints. Traditionally, nuclear re-actor design optimization relied on the experts’ experience and in-depth knowledge of each design step. It has inherent risks of the local optima. This paper proposes the multi-objective optimization framework consisting of surrogate modeling using regression algorithms and the metaheuristic search using a hybrid of genetic algorithm and simulated annealing. In our approach, multiple sets of design parameter inputs and objective value outputs from the Monte Carlo simulation code are evaluated using objective functions and constraints. In this step, correlation analysis and feasibility search based on the trust-region method are performed for better learning data sets to the surrogate model. Then, the Gaussian Process regression algorithm with suitable kernel and hyperparameters makes a surrogate model using the learning data. In the surrogate model, the optimal design parameters are searched through the iteration process of fitness evaluation, mating, average-bound crossover, wavelet mutation, and annealing. The fitness is evaluated by objective values processed with the weighted sum method based on the contribution to the objectives. If the error of optimal values and the surrogate model is below the threshold, the framework ends. If not, the additional sets of learning data are added based on the tentative optimal point, and a new surrogate model is built. The framework is applied to the TRU minimization problem using GT-HTR300 for single-batch core loading. Each TRU nuclides, initial loading of TRU and U-235 are set objectives, and 25.9% of TRU destruction rate and 58.7% of Pu-239 destruction rate with 3% of errors were achieved.
Subject
Multi-Objective OptimizationGenetic Algorithm
Surrogate Modeling
Reactor Design
Transuranics
HTR
Citation
Kajihara, Takanori (2022). Parametric Optimization of TRU Destruction Rates in HTR Cores Using Hybrid Metaheuristic Algorithm and Regression Methods. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /197147.