Long-term optimization of fuel loading pattern using genetic algorithms and simulated annealing

This paper describes Automatic Refueling Planning System (ARPS) for a nuclear power station using Genetic Algorithms (GA) and a Simulated Annealing (SA). ARPS has been developed and verified by applying to the Fugen nuclear power station (NPS), which is a 165MWe, heavy water-moderated, boiling light water-cooled, pressure tube-type reactor developed by JNC utilizing mainly uranium and plutonium mixed oxide (MOX) fuel. Fuel loading patterns have been managed independently in the Fugen NPS since the initial core. A planning of an adequate fuel loading pattern on each operational cycle needs one to two months even for expert core management engineers, for the reason that it has multi-objective optimization and nonlinear problems. In order to achieve the optimum fuel loading pattern and a fuel cost reduction, ARPS has been developed by JNC and CRC Solutions Corporation for the last five years. ARPS firstly generates several thousand fuel loading patterns with GA at random and secondly shuffles fuel assemblies with GA and SA by the computational procedure in an Engineering Work Station (EWS). Finally, the optimum core is estimated from the viewpoints of an effective multiplication factor, a power peaking factor, an average discharge fuel burn-up, a regional power and MOX fuel loading positions. GA based on the Darwinian theory of evolution is suited to multi-objective optimization. SA prevents local optimum solutions on a wide variety of combinatorial optimization problems. It has been verified that fuel loading patterns of the Fugen NPS evaluated by ARPS are equivalent to that of expert core management engineer. Furthermore, ARPS is capable of evaluating long-term fuel loading patterns of the Fugen NPS and applying real fuel loading patterns using Graphical User Interface (GUI). (author)