化工学报 ›› 2020, Vol. 71 ›› Issue (7): 3151-3164.doi: 10.11949/0438-1157.20191139

• 过程系统工程 • 上一篇    下一篇

基于主成分分析与支持向量机的热泵系统制冷剂泄漏识别研究

于仙毅1(),巫江虹1(),高云辉2   

  1. 1.华南理工大学机械与汽车工程学院,广东 广州 510641
    2.美的暖通设备有限公司,广东 佛山 528300
  • 收稿日期:2019-10-08 修回日期:2019-12-24 出版日期:2020-07-05 发布日期:2020-07-09
  • 通讯作者: 巫江虹 E-mail:2544168176@qq.com;pmjhwu@scut.edu.cn
  • 作者简介:于仙毅(1994—),男,硕士研究生,2544168176@qq.com
  • 基金资助:
    广州市科技计划项目(201804010287);广东省重点培育项目(2018B030308006)

Research on refrigerant leakage identification for heat pump system based on PCA-SVM models

Xianyi YU1(),Jianghong WU1(),Yunhui GAO2   

  1. 1.School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou 510641, Guangdong, China
    2.Midea Hvac Equipment Limited Company, Foshan 528300, Guangdong,China
  • Received:2019-10-08 Revised:2019-12-24 Online:2020-07-05 Published:2020-07-09
  • Contact: Jianghong WU E-mail:2544168176@qq.com;pmjhwu@scut.edu.cn

摘要:

为了研究热泵系统制冷剂泄漏识别的数据挖掘理论方法和实验验证,首先建立空气源热泵系统制冷剂泄漏实验台,进行热泵系统正常工况、干扰工况、泄漏工况的实验参数测试;其次,采用主成分分析法对测试数据进行特征提取处理,采用支持向量机对数据进行分类识别,建立了用于热泵系统的制冷剂泄漏识别的主成分分析-支持向量机模型,在二分类和多分类模式下验证了模型的性能,并研究了泄漏速率和不同故障工况对模型的影响。采用RefliefF特征选择算法对原始特征参数进行筛选,简化了识别模型的特征参数。研究结果表明:对于空气源热泵热水系统,PCA-SVM泄漏识别模型在多种验证集中对泄漏工况的识别准确度达100%,缓慢泄漏的诊断识别性能弱于快速泄漏,同一模型在不同故障诊断识别中性能不同,对系统运行影响轻微的故障诊断识别性能弱于其他故障。RefliefF特征选择方法将原始41个系统特征参数精简至10个特征参数,参数筛选优化后的泄漏识别模型识别精度也维持在较高水平,优化的泄漏识别模型更利于实际应用。

关键词: 泄漏识别, 支持向量机, 主成分分析, 特征选择, 热泵系统, 算法, 数学模型

Abstract:

To study the data mining theory method and experimental verification of the refrigerant leakage identification of the heat pump system, firstly establish an air source heat pump system refrigerant leakage test bench to test the experimental parameters of the heat pump system in normal working conditions, interference working conditions, and leakage working conditions. Then, principal component analysis (PCA) was used to process the experimental data, and support vector machine (SVM) was used to classify and identify the data. A leakage identification model based on PCA-SVM was established which verified in both two classification and multi classification model. The leakage rate and the influence of different fault conditions of the model was studied. RefliefF feature selection algorithm is used to screen the original feature parameters which simplify the feature parameters of the identification model. The results show that, for the air source heat pump water heater, the leakage identification model has a high identification of 100% in the leakage mode, and the slow leak diagnosis recognition performance of the weak in rapid leak, the same model in different fault diagnosis recognition performance is different, slight influence on the system s running fault diagnosis recognition performance is weaker than other malfunction. RefliefF feature selection method reduces the original 41 system characteristic parameters to 10 characteristic parameters. The identification accuracy of the leakage identification model after parameter screening and optimization is also maintained at a high level, the optimized leakage identification model is more conducive to practical application.

Key words: leakage identification, SVM, PCA, feature selection, heat pump, algorithm, mathematical modeling

中图分类号: 

  • TP 312

表1

制冷剂泄漏实验系统部件信息"

系统部件规格型号
制冷循环系统压缩机BSA645CV-R1EN型 R134a制冷剂
冷凝器U型套管式
节流阀丹佛斯TN2型热力膨胀阀
蒸发器(含小型风扇组)单流层微通道换热器
水路循环系统保温水箱30 L、?20 mm进出水口
水泵1个 自吸式磁力循环泵
水流量计1个 LWGY型涡轮流量计
制冷剂泄漏控制及收集部件手阀4个
开度阀4个
气体收集袋1个 20 L超高密封袋
测试部件热电偶若干 J型热电偶
压力变送器8个 0~0.6 MPa、0~4 MPa
电子秤100 g/0.02 g 15 kg/0.2 g
功率仪1个 HOPI型
安捷伦1台 34972型
计算机1台

图1

泄漏实验系统及测点布置示意图温度测点; 压力测点;1—压缩机;2—防爆气囊;3—可调开度阀;4—开度阀;5—水流量计;6—水箱;7—水泵;8—套管冷凝器;9—热力膨胀阀;10—蒸发器;11—风机;12—制冷剂气瓶;13—开度阀"

图2

制冷剂泄漏实验实物图"

表2

恒定水温热泵系统测试工况"

工况类型工况详情及引入方法
正常工况(normal)系统开机后,恒定30℃水温,系统稳定运行
泄漏工况(Refleak)以正常工况为基准,在其开机平稳运行一小段时间后,系统趋于稳定的一个时间点作为泄漏工况的开始点,开始控制泄漏口阀门开度
冷凝器脏污工况(ReduCF)调节冷凝器水泵,降低冷凝器水流量
蒸发器脏污工况(ReduEF)通过遮挡蒸发器,降低蒸发器换热面积
热力膨胀阀预紧力过小工况(loosTV)人为调松热力膨胀阀预紧弹簧
热力膨胀阀预紧力过大工况(ClosTV)人为调紧热力膨胀阀预紧弹簧
冷凝温度变化工况(IncrTW)(DecrTW)分别恒定6号大水箱内的水温为25℃和35℃

表3

特征变量名称符号及对应的系统表征含义"

序号

变量

符号

变量名称意义
1Tci1冷凝器进口温度1测点温度
2Tci2冷凝器进口温度2测点温度
3Tci3冷凝器进口温度3测点温度
4Tci4冷凝器进口温度4测点温度
5Tci5冷凝器进口温度5测点温度
6Tco1冷凝器出口温度1测点温度
7Tco2冷凝器出口温度2测点温度
8Tco3冷凝器出口温度3测点温度
9Tco4冷凝器出口温度4测点温度
10Tco5冷凝器出口温度5测点温度
11Tei1蒸发器进口温度1测点温度
12Tei2蒸发器进口温度2测点温度
13Tei3蒸发器进口温度3测点温度
14Tei4蒸发器进口温度4测点温度
15Tei5蒸发器进口温度5测点温度
16Teo1蒸发器出口温度1测点温度
17Teo2蒸发器出口温度2测点温度
18Teo3蒸发器出口温度3测点温度
19Teo4蒸发器出口温度4测点温度
20Teo5蒸发器出口温度5测点温度
21Tesu蒸发器风冷出口环温蒸发端环境温度
22Twat水箱水温冷凝端环境温度
23POcom压缩机耗功系统输入耗功
24Tesub过热度Teo1-Tgsat
25Tcsub过冷度Tlsat-Tco1
26P01压缩机出口压力压力测点
27P02冷凝器进口压力压力测点
28P03冷凝器出口压力压力测点
29P04节流阀入口压力压力测点
30P05节流阀出口压力压力测点
31P06蒸发器进口压力压力测点
32P07蒸发器出口压力压力测点
33P08压缩机进口压力压力测点
34ΔTcom压缩机进出口温差Tci1-Teo5
35ΔTcon冷凝器进出口温差Tci5-Tco1
36ΔTvel节流阀温差Tco5-Tei1
37ΔPcom压缩机进出口压差P01-P08
38ΔPcon冷凝器进出口压差P02-P03
39ΔPvel节流阀压差P04-P05
40ΔPeva蒸发器进出口压差P06-P07
41ΔHcon冷凝器进出口焓差Hci-Hco

图3

泄漏工况实验制冷剂泄漏速率"

表4

不同核函数类型的SVM模型表达式及其参数"

名称表达式参数
线性核kxi,xj=xiTxj
多项式核kxi,xj=xiTxjdd1为多项式次数
高斯核kxi,xj=exp-xi-xj22δ2δ>0为高斯核的带宽
拉普拉斯核kxi,xj=exp-xi-xjδδ>0
Sigmoid核kxi,xj=tanh?βxiTxj+θ

tanh为双曲正切函数,

β>0,θ<0

图4

基于PCA-SVM的制冷剂泄漏识别模型"

表5

二分类结果的混淆矩阵"

真实结果识别结果
泄漏非泄漏
泄漏TPFN
非泄露FPTN

表6

泄漏识别模型评级评价指标及其定义"

评价指标定义计算公式
按类性能命中率TPR对于给定类,发生且正确预测的样本占总发生样本的比率TPR=TPTP+FN
虚警率FPR对于给定类,没发生但被预测为发生的样本占没发生样本总数的比率FPR=FPFP+TN
总体性能准确率Acc正确分类数占总样本数的比率Acc=TP+FNTP+FP+FN+TN
错误分类率Mcr错误分类样本数占总样本数的比率Mcr=1-Acc

表7

泄漏特征的主成分分析结果"

主元编号特征值主元方差贡献率/%累计方差贡献率/%Tci1Tci2Tci3Tci4ΔPvelΔPevaΔHcon
115.8038.5538.550.0520.2200.246-0.135-0.2890.2490.112
29.6423.5162.060.0480.2130.253-0.143-0.0000.000-0.000
37.1217.3679.420.0480.2160.250-0.142-0.0000.000-0.000
43.127.6187.030.0450.2130.254-0.1450.0000.0000.000
52.375.7892.810.0340.2160.254-0.148-0.1210.3120.446
60.822.0194.82-0.1730.214-0.0720.0360.054-0.140-0.200
70.711.7496.56-0.2080.157-0.0520.036-0.000-0.000-0.000

表8

不同核函数类型的SVM模型信息"

SVM模型编号名称意义
1Linear SVM线性核函数
2Quadratic SVM二次多项式核函数 d=2
3Cubic SVM三次多项式核函数 d=3
4Fine Gaussian SVM精细高斯核函数 δ20
5Medium Gaussian SVM中位高斯核函数 0<δ2<
6Coarse Gaussisn SVM粗糙高斯核函数 δ2

图5

泄漏识别模型训练过程识别准确率变化"

图6

泄漏识别模型训练过程识别命中率和虚警率变化"

表9

泄漏/非泄漏模式下PCA-SVM识别模型及性能"

CPVa主元组合SVM核函数Acc/%Mcr/%TPR/%FPR/%测试集混淆矩阵
87.02%[1,2,3,4]Fine Gaussian SVM100099.5020510436

图7

不同模型在各个故障的诊断识别准确率"

表10

不同模型在各个故障的诊断识别的混淆矩阵"

工况Model-oModel-pca4

Refleak

normal

ReduCF

ReduEF

loosTV

ClosTV

IncrTW

DecrTW

25800000004225000000402300000730250000150007600081000520030000042030000005425440000001228000000002600000060281000030088000510015400000000460000000057
工况Model-pca5Model-pca6

Refleak

normal

ReduCF

ReduEF

loosTV

ClosTV

IncrTW

DecrTW

25710000003224020000002600000130281000110089000010015900000000460000000057256200000042230200001223000004130171000510085000410005600110000440100000056
工况Model-pca7

Refleak

normal

ReduCF

ReduEF

loosTV

ClosTV

IncrTW

DecrTW

256200000010216030000412100000814013000014100760001300004800600000460200000055

图8

四种模型各泄漏故障的诊断识别准确率对比"

图9

四种模型在相同数据集中的各泄漏诊断识别准确率对比"

表11

四种模型在相同数据集的各泄漏诊断识别结果混淆矩阵"

工况Model-pca4Model-pca5Model-pca6Model-pca7

Refleak_slow

Refleak_fast

normal

9114021503045399690215210054391122514550154280121331391300543

图10

RefliefF特征选择权重结果"

表12

RefliefF特征选择前后的PCA-SVM泄漏识别模型及性能结果对比"

方式序号主元组合SVM核函数Acc/%Mcr/%TPR/%FPR/%测试集混淆矩阵
PCA-SVM1[1,2,3,4]Fine Gaussian SVM100099.5020510436
2[12,3,4,5]100099.5020510436
3[1,2,3,4,5,6]99.10.997.60.220151435

RefliefF

PCA-SVM

4[1r,2r,3r,4r]Fine Gaussian SVM97.82.297.11.820068428
5[1r,2r,3r,4r,5r]97.42.695.61.819798428
6[1r,2r,3r,4r,5r,6r]97.42.693.70.9193134432
1 Mohanraj M, Muraleedharan C, Jayaraj S. A review on recent developments in new refrigerant mixtures for vapour compression-based refrigeration, air-conditioning and heat pump units[J]. International Journal of Energy Research, 2011, 35(8): 647-669.
2 王晓明. 制冷系统故障先兆分析和故障预报技术研究[D]. 上海: 上海交通大学, 1998.
Wang X M. Analysis of refrigeration system failure predication and failure prediction technology [D]. Shanghai: Shanghai Jiaotong University, 1998.
3 Koronaki I P, Cowan D, Maidment G, et al. Refrigerant emissions and leakage prevention across Europe-results from the RealSkillsEurope project[J]. Energy, 2012, 45(1): 71-80.
4 Francis C, Maidment G, Davies G. An investigation of refrigerant leakage in commercial refrigeration[J]. International Journal of Refrigeration, 2017, 74: 12-21.
5 Cowan D, Maidment G, Churchyard B, et al. Maintenance and long-term operation of supermarkets and minimizing refrigerant leakage[M]// Sustainable Retail Refrigeration. New York: John Wiley & Sons Ltd., 2015: 14-20.
6 Chantant M, Lambert R, Gargiulo L, et al. Leak tightness tests on actively cooled plasma facing components: lessons learned from Tore Supra experience and perspectives for the new fusion machines[J]. Fusion Engineering & Design, 2015, 98: 92-96.
7 Tassou S A, Grace I N. Fault diagnosis and refrigerant leak detection in vapour compression refrigeration systems[J]. International Journal of Refrigeration, 2005, 28(5): 680-688.
8 任能. 制冷系统故障检测、诊断及预测研究[D]. 上海: 上海交通大学, 2008.
Ren N. Fault detection, diagnosis and prediction of refrigeration system [D]. Shanghai: Shanghai Jiaotong University, 2008.
9 Beshr M, Aute V, Sharma V, et al. A comparative study on the environmental impact of supermarket refrigeration systems using low GWP refrigerants[J]. International Journal of Refrigeration, 2015, 56: 154-164.
10 Coulomb D. Refrigeration and cold chain serving the global food industry and creating a better future: two key IIR challenges for improved health and environment[J]. Trends in Food Science & Technology, 2008, 19(8): 409-417.
11 郭军峰. 汽车空调性能衰减研究[D].重庆: 重庆大学, 2012.
Guo J F. Research on performance attenuation of automobile air conditioner [D]. Chongqing: Chongqing University, 2012.
12 Yoo J W, Hong S B, Kim M S. Refrigerant leakage detection in an EEV installed residential air conditioner with limited sensor installations[J]. International Journal of Refrigeration, 2017, 78: 157-165.
13 Yu Y, Woradechjumroen D, Yu D. A review of fault detection and diagnosis methodologies on air-handling units[J]. Energy and Buildings, 2014, 82: 550-562.
14 Grace I N, Datta D, Tassou S A. Sensitivity of refrigeration system performance to charge levels and parameters for on-line leak detection[J]. Applied Thermal Engineering, 2005, 25(4): 557-566.
15 Grace I N, Datta D, Tassou S A. Sensitivity of refrigeration system performance to charge levels and parameters for on-line leak detection[J]. Applied Thermal Engineering, 2005, 25(4): 557-566.
16 Yoon S H, Payne W V, Domanski P A. Residential heat pump heating performance with single faults imposed[J]. Applied Thermal Engineering, 2011, 31(5): 765-771.
17 Han H, Gu B, Wang T, et al. Important sensors for chiller fault detection and diagnosis (FDD) from the perspective of feature selection and machine learning[J]. International Journal of Refrigeration, 2011, 34(2): 586-599.
18 Han H, Cao Z, Gu B, et al. PCA-SVM-based automated fault detection and diagnosis (AFDD) for vapor-compression refrigeration systems[J]. HVAC&R Research, 2010, 16(3): 295-313.
19 齐咏生, 张海利, 王林, 等.基于MSPCA-KECA的冷水机组故障监测及诊断[J].化工学报, 2017, 68(4): 1499-1508.
Qi Y S, Zhang H L, Wang L, et al. Fault monitoring and diagnosis of chiller based on MSPCA-KECA [J]. CIESC Journal, 2017, 68(4): 1499-1508.
20 韩华, 谷波, 任能. 基于主元分析与支持向量机的制冷系统故障诊断方法[J]. 上海交通大学学报, 2011, (9): 108-114+126.
Han H, Gu B, Ren N. Fault diagnosis method of refrigeration system based on principal component analysis and support vector machine [J]. Journal of Shanghai Jiaotong University, 2011, (9): 108-114+126.
21 洪迎春. 基于多变量统计分析的制冷系统故障检测与诊断[D]. 上海: 上海交通大学, 2012.
Hong Y C. Fault detection and diagnosis of refrigeration system based on multivariate statistical analysis [D]. Shanghai: Shanghai Jiaotong University, 2012.
22 梁晴晴, 韩华, 崔晓钰, 等. 基于主元分析-概率神经网络的制冷系统故障诊断[J]. 化工学报, 2016, 67(3): 1022-1031.
Liang Q Q, Han H, Cui X Y, et al. Fault diagnosis of refrigeration system based on PCA-PNN [J]. CIESC Journal, 2016, 67(3): 1022-1031.
23 王路瑶, 吴斌, 杜志敏, 等.基于长短期记忆神经网络的数据中心空调系统传感器故障诊断[J].化工学报, 2018, 69: 252-259.
Wang L Y, Wu B, Du Z M, et al. Sensor fault diagnosis of air conditioning system in data center based on neural network of long and short term memory [J]. CIESC Journal, 2018, 69: 252-259.
24 韩华. 基于顺序集成方法的制冷系统故障检测与诊断研究[D].上海: 上海交通大学, 2012.
Han H. Study on fault detection and diagnosis of refrigeration system based on sequential integration method [D]. Shanghai: Shanghai Jiaotong University, 2012.
25 赵旭, 阎威武, 邵惠鹤, 等. 基于核Fisher判别分析方法的非线性统计过程监控与故障诊断[J]. 化工学报, 2007, 58(4): 951-956.
Zhao X, Yan W W, Shao H H, et al. Nonlinear statistical process monitoring and fault diagnosis based on nuclear Fisher discriminant analysis [J]. Journal of Chemical Industry and Engineering(China), 2007, 58(4): 951-956.
26 周志华, 王珏. 机器学习及其应用2009[M]. 北京: 清华大学出版社, 2009.
Zhou Z H, Wang J. Machine Learning and Its Application 2009[M]. Beijing: Tsinghua University Press, 2009.
27 Valle S, Li W, Qin S J. Selection of the number of principal components: the variance of the reconstruction error criterion with a comparison to other methods[J]. Industrial & Engineering Chemistry Research, 1999, 38(11): 4389-4401.
28 Jos B, Draper B A. Feature selection from huge feature sets[C]// Proceedings. Eighth IEEE International Conference on Computer Vision. IEEE, 2001.
29 Robnik M, Kononenko I. Theoretical and empirical analysis of ReliefF and RReliefF[J]. Machine Learning, 2003, 53(1/2): 23-69.
30 Deng X, Li Y, Weng J, et al. Feature selection for text classification: a review[J]. Multimedia Tools and Applications, 2018, 78: 3797-3816.
[1] 马坤茹, 李雪峰, 李思琦, 高翠娟. 新型太阳能/空气能直膨式热泵与空气源热泵供热性能对比[J]. 化工学报, 2020, 71(S1): 375-381.
[2] 张洁, 庞丽萍, 曲洪权, 王天博. 基于随机配置网络的机载电子吊舱多工况热模型[J]. 化工学报, 2020, 71(S1): 441-447.
[3] 高学金, 刘腾飞, 徐子东, 高慧慧, 于涌川. 基于循环自动编码器的间歇过程故障监测[J]. 化工学报, 2020, 71(7): 3172-3179.
[4] 杨辉, 代文豪, 陆荣秀, 朱建勇. 基于分离系数校正的稀土萃取流程模拟[J]. 化工学报, 2020, 71(7): 3180-3190.
[5] 李炜, 王晓明, 蒋栋年, 李亚洁, 梁成龙. 基于SHPSO-GA-BP的成品汽油调和中加氢汽油组分辛烷值的预测[J]. 化工学报, 2020, 71(7): 3191-3200.
[6] 邓睿渠, 汪林正, 张睿智, 罗永浩. 基于粒子群算法的生活垃圾高热值成分热解动力学特性研究[J]. 化工学报, 2020, 71(7): 3238-3246.
[7] 魏伟, 李秀财, 孙奉仲. 超细飞灰对烟气酸露点与酸凝结的影响研究[J]. 化工学报, 2020, 71(7): 3258-3265.
[8] 蔡涛, 杨博, 李宏光. 基于时延挖掘模糊时间认知图的化工过程多变量时序预测方法[J]. 化工学报, 2020, 71(3): 1095-1102.
[9] 杜宇浩, 阎高伟, 李荣, 王芳. 基于局部线性嵌入的测地线流式核多工况软测量建模方法[J]. 化工学报, 2020, 71(3): 1278-1287.
[10] 耿志强, 曾荣甫, 徐圆, 韩永明, 顾祥柏. 融合灰狼优化算法在工控系统入侵检测中的应用[J]. 化工学报, 2020, 71(3): 1080-1087.
[11] 耿志强, 毕帅, 王尊, 朱群雄, 韩永明. 基于改进NSGA-Ⅱ算法的乙烯裂解炉操作优化[J]. 化工学报, 2020, 71(3): 1088-1094.
[12] 谢雨珩, 李智, 杨明磊, 杜文莉. 基于自适应采样算法的芳烃异构化代理模型[J]. 化工学报, 2020, 71(2): 688-697.
[13] 唐振浩,张宝凯,曹生现,王恭,赵波. 基于多模型智能组合算法的锅炉炉膛温度建模[J]. 化工学报, 2019, 70(S2): 301-310.
[14] 刘磊磊,夏新林,侯凌霄,孙创,陈学. 基于二维渗流方程的微纳孔隙材料渗透率反演求解[J]. 化工学报, 2019, 70(S2): 343-348.
[15] 王斯民, 孙利娟, 宋晨, 张早校, 文键. 螺旋扁管折流杆换热器壳侧性能多目标优化研究[J]. 化工学报, 2019, 70(9): 3353-3362.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!