摘要
为了保证从核电站大量数据中有效地挖掘信息以及故障下运行状态的智能表征,本文提出一种基于粒子群优化和最小二乘支持向量机的系统级故障程度评估方法,用于完善故障诊断系统的功能。针对最小二乘支持向量机算法的超参数选取对于回归精度影响较大问题,应用基于粒子群优化算法借助智能搜索策略来优化模型的超参数。基于最优超参数的回归模型能够提取系统级参数间的约束关系,以进行实时故障程度的评估。性能测试表明:采用提出的方法能够有效评估核电站系统级故障的程度,相较于粒子群优化-支持向量机以及最小二乘支持向量机算法具有更高的回归精度,且抗噪性能良好,保证了故障诊断系统的精度及可靠性。
To ensure the effective mining of information from a large amount of data and the intelligent representation of the operating state of nuclear power plant(NPP)under failure,this paper proposes a system-level failure severity evaluation method based on particle swarm optimization(PSO)and least square support vector machine(LSSVM)to improve the function of a fault diagnosis system.In view of the problem that the hyperparameter selection of the LSSVM algorithm has a great influence on the regression accuracy,the PSO algorithm is employed to optimize the LSSVM model’s hyperparameters using an intelligent search strategy.The regression model based on the optimal hyperparameters can extract the constraint relationship between system-level parameters for real-time fault severity assessments.Various performance tests show that the proposed method can effectively evaluate the severity of system-level failures of NPPs.Compared with the PSO-SVM and LSSVM algorithms,the proposed method has higher regression accuracy and better noise resistance,which ensures the accuracy and reliability of the fault diagnosis system.
作者
王志超
夏虹
彭彬森
朱少民
WANG Zhichao;XIA Hong;PENG Binsen;ZHU Shaomin(Key Laboratory of Nuclear Safety and Advanced Nuclear Energy Technology, Harbin Engineering University, Harbin 150001, China;Fundamental Science on Nuclear Safety and Simulation Technology Laboratory, Harbin Engineering University, Harbin 150001, China)
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2021年第12期1748-1753,共6页
Journal of Harbin Engineering University
基金
国家自然科学基金项目(U21B2083)
国防科技工业核动力技术创新中心项目(HDLCXZX-2021-ZH-019)
中央高校基本科研业务费专项资金资助项目(3072021GIP1503).
关键词
核动力装置
故障程度评估
最小二乘-支持向量机
粒子群优化算法
运行支持
回归模型
优化算法
数据驱动
nuclear power plants
failure severity evaluation
least square support vector machine
particle swarm optimization
operation support
regression model
optimization algorithm
data-drive
