Research On Fuel Reloading Optimization Method Of Block-type HTGRs Based On Hybrid Teaching-Learning Genetic Algorithm

By solving the problems of nuclear reactor fuel reloading optimization,the optimal core configuration of fuel assemblies could be obtained to reduce the nuclear fuel cycle cost.However,this problem is a complex combinational optimization problem with a very huge solution space,which is difficult to be solved by enumeration method.With the development of computer science and technology,some optimization algorithms are gradually applied to solve the problems of fuel reloading optimization.These optimization algorithms could continuously update the core configurations of fuel assemblies and combine the core physics calculation programs to evaluate them,which could obtain better solutions in a limited time.Among them,most optimization algorithms have certain limitations when solving the problems of fuel reloading optimization.For example,the convergence speed of genetic algorithm is relatively slow,while teaching-learning based optimization algorithms is likely to fall into local optimal.In order to get an algorithm with better optimization ability,the hybrid teaching-learning genetic algorithm is developed in this paper which is based on the teachinglearning based optimization algorithm,combing coding,crossover and mutation these three operators in genetic algorithm.At the same time,the “Students” individuals in teachinglearning based optimization algorithm are further divided into top students,ordinary students and poor students;and the two calculation phases “Teacher phase” and “Learner phase” are developed to “Teacher phase”,“Discussion phase” and “Self-study phase” these three phase.For testing the optimization ability of the hybrid algorithm,taking the 1/6 core of block-type HTGRs as objects,the hybrid teaching-learning genetic algorithm is used for fuel reloading optimization calculation.The calculation results show that this hybrid algorithm could effectively overcome the defect of premature convergence for teaching-learning based optimization algorithm.At the same time,comparing to genetic algorithm,the hybrid teachinglearning genetic algorithm could further accelerate the convergence speed.However,in traditional calculation method,the calculation time of core physics is more than 40 hours,approximately accounting for 99.8% of the total time,which affects the optimization process of optimization algorithms.At the same time,based on the alternative model,a novel calculation method for fuel reloading optimization is developed to quickly evaluate the solutions of the problems of fuel reloading optimization.Using the alternative models of core physics calculation,taking the core configurations of fuel assemblies as inputs,the evaluation results could be obtained by simple matrix calculation,thus avoiding the complicated transportation calculation,burnup calculation and other processes.For testing the performance of this alternative model in the actual calculation for fuel reloading optimization,the model is combined with hybrid teaching-learning genetic algorithm to execute the calculation of fuel reloading optimization.The results show that in novel calculation method,the hybrid teaching-learning genetic algorithm has better performance than teaching-learning based optimization algorithm and genetic algorithm.In addition,the testing accuracy of the alternative model in novel calculation method is also excellent,reaching 0.74.Finally,comparing to traditional calculation method of fuel reloading optimization,the quality of optimization results by novel calculation method is similar to that of traditional calculation method.Hovewer,the calculation time of novel calculation method is only 0.7% of the traditional calculation method,which greatly speeds up the calculation speed.