我的订单 | 我的收藏 | 加入收藏 | 返回首页
暖通空调杂志社>期刊目次>2018年>第9期

基于粒子群优化最小二乘支持向量机的离心式制冷机故障诊断

Fault diagnosis of centrifugal chillers based on particle swarm optimization-least squares support vector machine

卿 红 韩 华 崔晓钰 范雨强
上海理工大学

摘要:

针对制冷系统传统故障诊断正确率低的问题,引入最小二乘支持向量机(LSSVM)算法用于制冷系统故障诊断。在LSSVM模型基础上,结合粒子群优化(PSO)得到PSO-LSSVM模型,利用特征选择方法优化得到LSSVM8模型,利用组合方法得到PSO-LSSVM8模型。分析比较了4种模型的诊断性能。结果表明:PSO-LSSVM模型、LSSVM8模型均可改善基于LSSVM模型的制冷系统故障诊断性能,尤其是对于制冷剂泄漏/充注量不足故障,准确率分别提高1.04%,1.24%;PSO-LSSVM8模型比采用单种优化方法的诊断模型具有更好的诊断性能,可克服人为选择参数的盲目性,在全局优化与收敛速度方面具有较大优势,应用于制冷系统故障诊断具有较好的可行性。

关键词:制冷系统,故障诊断,最小二乘支持向量机,粒子群算法,优化

Abstract:

Aiming at the low accuracy rate of traditional fault diagnosis in refrigeration system, presents the least squares support vector machine (LSSVM) algorithm for the fault diagnosis field of refrigeration system. Based on the LSSVM model, obtains the PSO-LSSVM model by combining the particle swarm optimization (PSO) algorithm, the LSSVM8 model by the feature selection method, and the PSO-LSSVM8 model by the combination method. Analyses and compares the diagnostic performance of the four models. The results show that the PSO-LSSVM model and LSSVM8 model can improve the performance of the fault diagnosis of the refrigeration system based on the LSSVM model in different aspects, especially for the failure of the refrigerant leakage/filling, and the accuracy rate increases by 1.04%, 1.24% respectively, and the PSO-LSSVM8 model has better diagnostic performance than the single optimization method, it can overcome the blindness of artificial parameters selection, with great advantages in the comprehensive optimization and convergence speed, which has good feasibility in the fault diagnosis of the refrigeration system.

Keywords:refrigerationsystem,faultdiagnosis,leastsquaressupportvectormachine,particleswarmalgorithm,optimization

    你还没注册?或者没有登录?这篇期刊要求至少是本站的注册会员才能阅读!

    如果你还没注册,请赶紧点此注册吧!

    如果你已经注册但还没登录,请赶紧点此登录吧!