Development and Utilization of mathematical Optimization in Advanced Fuel Cycle Systems Analysis

Turinsky, Paul ; Hays, Ross

关键词: ALGORITHMS;    ANNEALING;    CAPACITY;    ELECTRIC POWER;    FUEL CYCLE;    NUCLEAR FUELS;    NUCLEAR POWER;    OPTIMIZATION;    PROLIFERATION;    RECYCLING;    SIMULATION;    SYSTEMS ANALYSIS;    WASTE DISPOSAL;   

DOI  :  10.2172/1024390
RP-ID  :  DOE/ID/14734
PID  :  OSTI ID: 1024390
Others  :  TRN: US201119%%348

美国|英语

来源: SciTech Connect

【 摘 要 】

Over the past sixty years, a wide variety of nuclear power technologies have been theorized, investigated and tested to various degrees. These technologies, if properly applied, could provide a stable, long-term, economical source of CO2-free electric power. However, the recycling of nuclear fuel introduces a degree of coupling between reactor systems which must be accounted for when making long term strategic plans. This work investigates the use of a simulated annealing optimization algorithm coupled together with the VISION fuel cycle simulation model in order to identify attractive strategies from economic, evironmental, non-proliferation and waste-disposal perspectives, which each have associated an objective function. The simulated annealing optimization algorithm works by perturbing the fraction of new reactor capacity allocated to each available reactor type (using a set of heuristic rules) then evaluating the resulting deployment scenario outcomes using the VISION model and the chosen objective functions. These new scenarios, which are either accepted or rejected according the the Metropolis Criterion, are then used as the basis for further perturbations. By repeating this process several thousand times, a family of near-optimal solutions are obtained. Preliminary results from this work using a two-step, Once-through LWR to Full-recycle/FRburner deployment scenario with exponentially increasing electric demand indicate that the algorithm is capable of nding reactor deployment pro les that reduce the long-term-heat waste disposal burden relative to an initial reference scenario. Further work is under way to re ne the current results and to extend them to include the other objective functions and to examine the optimization trade-o s that exist between these di erent objectives.