论文部分内容阅读
为解决突发核电事故后使用机理模型预测放射性液态流出物迁移扩散,需长时间迭代计算的不足,提出了一种新型混合神经网络模型,该模型耦合了描述液态流出物在受纳水体中迁移扩散的组分输运方程和神经网络模型,采用并行多种群混合进化粒子群算法计算神经网络权值与阈值.论文以湖北咸宁大畈核电站受纳水体富水水库为研究对象,对事故工况下长半衰期核素迁移扩散进行模拟预测,研究结果表明有一定物理基础的神经网络模型是一种有效、可行的预测模型,预测结果与机理模型的模拟输出拟合度较好,新模型较传统的黑箱神经网络模型以及基于单调型先验知识的神经网络模型具有更强的泛化性能改善能力.
In order to solve the problem of using the mechanism model to predict the migration and diffusion of radioactive liquid effluent after a sudden nuclear accident, a new hybrid neural network model is proposed to solve the shortcomings of long-term iterative calculation. The model is coupled with a new hybrid neural network model that describes the migration of liquid effluent in receiving water Diffusion component transport equation and neural network model, the parallel multi-population hybrid evolutionary particle swarm optimization algorithm is used to calculate the neural network weights and thresholds.In this paper, the water-absorbing reservoir in Xianning Daban Nuclear Power Station in Hubei Province is taken as the research object, The results show that the neural network model with a certain physical basis is an effective and feasible prediction model, and the fitting result between the prediction result and the mechanism model is better. The new model is more traditional The black-box neural network model and the neural network model based on the prior knowledge of the monotonous model have stronger generalization performance improvement ability.