基于改进线性学习算法的核动力系统事故诊断研究
doi: 10.13832/j.jnpe.2020.01.0134
Research on Fault Diagnosis of Nuclear Power System Based on Improved Linear Learning Algorithm
-
摘要: 为解决核动力系统事故类型多样且故障严重程度难以确定的问题,在传统线性模型的基础上引入层级结构和嵌套结构,并选用支持向量机分类模型作为结构内的诊断模型;采用线性学习实现计算结果的融合,通过分析事故运行过程和机理选取单个分类模型的训练样本,并确定对应类别事故的有效识别区域及敏感参数。结果表明,本文提出的事故诊断框架的识别准确率达到99%以上,可为大型系统的事故诊断提供参考。Abstract: Because the types of nuclear power system accidents are various and the severity of accidents is difficult to determine, the hierarchical structure and nested structure are introduced on the basis of traditional linear model. The support vector machine classification model is selected as the diagnosis model in the structure, and the linear learning merges the results. By analyzing the operation process and mechanism of the accident, the effective identification area and sensitive parameters of the corresponding type of accident are determined. The results show that the final recognition accuracy rate is more than 99%, and it can provide reference for accident diagnosis in large-scale systems.
点击查看大图
计量
- 文章访问数: 170
- HTML全文浏览量: 19
- PDF下载量: 3
- 被引次数: 0