适用于核动力设备故障诊断的改进粒子群优化算法
2017-10-17李铁萍周国强田欣鹭
刘 锐, 李铁萍, 周国强, 田欣鹭
(1. 环境保护部核与辐射安全中心, 北京 100082; 2. 交通运输部规划研究院, 北京 100028)
适用于核动力设备故障诊断的改进粒子群优化算法
刘 锐1, 李铁萍1, 周国强2, 田欣鹭1
(1. 环境保护部核与辐射安全中心, 北京 100082; 2. 交通运输部规划研究院, 北京 100028)
提出一种基于改进粒子群优化(IPSO)算法的核动力设备故障诊断方法.利用已知核动力设备故障征兆集合,选用概率因果模型求解具有最大后验概率的故障集合;在传统粒子群优化(PSO)算法的基础上,利用佳点集原理对PSO算法进行初始化,优化了粒子群的初始化范围;借助自适应调整的惯性权重法,避免PSO算法未成熟收敛,加快了收敛速度.最后通过算例证明该方法的有效性.结果表明:基于改进粒子群优化算法的概率因果模型不受故障样本的限制,具有较好的通用性,且模型故障诊断精度较高、寻优速度快.
核动力设备; 故障诊断; 粒子群优化算法; 佳点集
Abstract: An improved particle swarm optimization (IPSO) algorithm was proposed for fault diagnosis of nuclear power systems. By using the known symptom sets of nuclear power faults, and with the probabilistic causal model, the IPSO algorithm was introduced to solve the fault sets with maximum a posteriori probability; based on traditional PSO algorithm, the principle of good point set was used to initialize the range of PSO; by adaptive adjustment of inertia weight, the premature convergence of PSO was avoided, and the convergence speed was accelerated. Finally, the validity of the method was demonstrated by examples. Results show that the probabilistic causal model based on IPSO algorithm is not limited by fault samples, which therefore has good versatility, with high precision in fault diagnosis and high speed in optimization.
Key words: nuclear power equipment; fault diagnosis; particle swarm optimization algorithm; good point set
故障诊断系统作为先进核动力仪表控制系统的重要组成部分,是运行支持系统的关键功能之一.目前,核动力设备故障诊断的主要方法是依靠运行人员的经验或穷举方法进行诊断排除[1].为保障核动力仪表控制系统的安全经济运行,当核动力设备发生故障时,需及时准确地识别出来.但是相对于其他动力设备,核动力设备具有故障数据采集困难、样本较少,甚至没有完整有效样本的特点,且其相互关联导致故障机理复杂,给核动力……
