基于数据增强的图卷积神经网络的浓度预测方法

罗丽平; 曾腾

doi:null

您当前的位置：

首页 >

文章列表页 >

基于数据增强的图卷积神经网络的浓度预测方法

更新时间：2026-03-03

- 基于数据增强的图卷积神经网络的浓度预测方法
- A Data-Augmentation-Based Graph Convolutional Neural Network for Concentration Prediction
- 物联网学报 2026年
- 作者机构：
  
  1.广西民族大学物理与电子信息学院，广西南宁 530006
  2.多模态信息智能感知处理与应用广西高校工程研究中心, 广西南宁 530006
  3.广西智语人形机器人重点实验室, 广西南宁 530006
  4.广西智能视觉协作机器人工程研究中心, 广西南宁 530006
- 作者简介：
  
  罗丽平，luoliping@gxmzu.edu.cn
- 基金信息：
  
  国家自然科学基金(62561006);广西科技重大专项(桂科AA23073006);广西重点研发计划(桂科AB25069215)
- DOI：
  中图分类号： TP212.9;TP183;X832
- 收稿：2025-10-23，
  
  修回：2026-01-23，
  
  录用：2026-02-09，
- 稿件说明：
移动端阅览
罗丽平, 曾腾. 基于数据增强的图卷积神经网络的浓度预测方法[J/OL]. 物联网学报, 2026.

LUO Liping, ZENG Teng. A Data-Augmentation-Based Graph Convolutional Neural Network for Concentration Prediction[J/OL]. Chinese Journal on Internet of Things, 2026.
罗丽平, 曾腾. 基于数据增强的图卷积神经网络的浓度预测方法[J/OL]. 物联网学报, 2026. DOI：

LUO Liping, ZENG Teng. A Data-Augmentation-Based Graph Convolutional Neural Network for Concentration Prediction[J/OL]. Chinese Journal on Internet of Things, 2026. DOI：

摘要

为解决电容式浓度预测在样本稀疏与噪声干扰下精度下降的问题，本研究提出一种基于数据增强的图卷积神经网络（GCN

Graph Convolutional Network）的浓度预测方法。对射频识别（RFID

Radio Frequency Identification）无线传感系统采集的实验数据，先采用插值提高数据密度，再通过卡尔曼滤波从观测序列中提取隐变量并与原特征拼接以扩充维度；然后构建含深度残差连接的GCN，学习电容与浓度之间的非线性映射，并采用不同图卷积方法进行性能评估。实验结果表明，该方法浓度预测的平均相对误差（MRE

Mean Relative Error）为2.35%，明显低于线性递减权重的粒子群优化等现有方法。此外，在不同的图卷积方法下，其预测性能仍保持稳定。由此可见，本研究提出的基于数据增强的图卷积神经网络架构，能有效解决数据稀疏、含噪情况下盐溶液浓度检测的准确性与鲁棒性问题，具有向多盐体系与复杂工况推广应用的潜力。

Abstract

To address the accuracy degradation of capacitive concentration prediction under sparse samples and noise interference

this study proposes a concentration prediction method based on a data-augmented Graph Convolutional Network (GCN). Experimental data collected from a Radio Frequency Identification (RFID) wireless sensing system are first processed through interpolation to increase data density. A Kalman filter is then applied to extract hidden variables from the observation sequence

which are concatenated with the original features to expand the feature dimension. A GCN with deep residual connections is constructed to learn the nonlinear mapping between capacitance and concentration. Different graph convolution strategies are further evaluated for performance comparison. Experimental results show that the proposed method achieves a mean relative error (MRE) of 2.35%

which is significantly lower than existing approaches such as the linearly decreasing weight particle swarm optimization. Moreover

the prediction accuracy remains stable across various convolution strategies. These results demonstrate that the proposed data-augmented GCN effectively enhances the accuracy and robustness of salt solution concentration detection under sparse and noisy data conditions

with potential for extension to multi-salt systems and complex environments.

关键词

Keywords

references

CRINI G , LICHTFOUSE E . Advantages and disadvantages of techniques used for wastewater treatment [J ] . Environmental Chemistry Letters , 2019 , 17 ( 1 ): 145 - 155 .

WEN J , DONG H R , ZENG G M . Application of zeolite in removing salinity/sodicity from wastewater: a review of mechanisms, challenges and opportunities [J ] . Journal of Cleaner Production , 2018 , 197 : 1435 - 1446 .

PERELONIA K B S , BENITEZ K C D , BANICOD R J S , et al . Validation of an analytical method for the determination of cadmium, lead and mercury in fish and fishery resources by graphite furnace and cold vapor atomic absorption spectrometry [J ] . Food Control , 2021 , 130 : 108363 .

ZHOU C , YU W T , YANG Y , et al . A new ion concentration detection method based on spectral matching and fusion for zinc leaching liquid [J ] . IEEE Transactions on Instrumentation and Measurement , 2022 , 72 : 1 - 12 .

MONTERROSO S C C , CARAPUÇA H M , SIMÃO J E J , et al . Optimisation of mercury film deposition on glassy carbon electrodes: evaluation of the combined effects of pH, thiocyanate ion and deposition potential [J ] . Analytica Chimica Acta , 2004 , 503 ( 2 ): 203 - 212 .

QIAN X Q , LI Z , MENG Z Z , et al . Flexible RFID tag for sensing the total minerals in drinking water via smartphone tapping [J ] . IEEE Sensors Journal , 2021 , 21 ( 21 ): 24749 - 24758 .

MAKAROVAITE V , HILLIER A J R , HOLDER S J , et al . Passive wireless UHF RFID antenna label for sensing dielectric properties of aqueous and organic liquids [J ] . IEEE Sensors Journal , 2019 , 19 ( 11 ): 4299 - 4307 .

XIE S , MA C L , FENG R H , et al . Wireless glucose sensing system based on dual-tag RFID technology [J ] . IEEE Sensors Journal , 2022 , 22 ( 13 ): 13632 - 13639 .

FENG R H , XIANG X W , XIE S , et al . Sensing system for mixed inorganic salt solution based on improved double label coupling RFID [J ] . IEEE Sensors Journal , 2023 , 23 ( 12 ): 13565 - 13574 .

MAZZARACCHIO V , FIORE L , NAPPI S , et al . Medium-distance affordable, flexible and wireless epidermal sensor for pH monitoring in sweat [J ] . Talanta , 2021 , 222 : 121502 .

KHAN M R R , KHALILIAN A , SEO J , et al . Highly reliable passive RFID-based inductor-capacitor sensory system strengthened by solvatochromism for fast and wide-range lactate detection [J ] . IEEE Sensors Journal , 2022 , 22 ( 12 ): 12228 - 12236 .

AKINOSHO T D , OYEDELE L O , BILAL M , et al . A scalable deep learning system for monitoring and forecasting pollutant concentration levels on UK highways [J ] . Ecological Informatics , 2022 , 69 : 101609 .

LI L , ZHANG R X , SUN J D , et al . Monitoring and prediction of dust concentration in an open-pit mine using a deep-learning algorithm [J ] . Journal of Environmental Health Science and Engineering , 2021 , 19 ( 1 ): 401 - 414 .

JIA M S , GABRYS B , MUSIAL K . A network science perspective of graph convolutional networks: a survey [J ] . IEEE Access , 2023 , 11 : 39083 - 39122 .

FU S C , LIU W F , ZHOU Y C , et al . HpLapGCN: hypergraph p-Laplacian graph convolutional networks [J ] . Neurocomputing , 2019 , 362 : 166 - 174 .

JIN G Y , LIANG Y X , FANG Y C , et al . Spatio-Temporal Graph Neural Networks for Predictive Learning in Urban Computing: A Survey [J ] . IEEE Transactions on Knowledge and Data Engineering , 2024 , 36 ( 10 ): 5388 - 5408 .

WU Z H , PAN S R , CHEN F W , et al . A Comprehensive Survey on Graph Neural Networks [J ] . IEEE Transactions on Neural Networks and Learning Systems , 2020 , 32 ( 1 ): 4 - 24 .

LI G H , MÜLLER M , QIAN G C , et al . DeepGCNs: Making GCNs Go as Deep as CNNs [J ] . IEEE Transactions on Pattern Analysis and Machine Intelligence , 2023 , 45 ( 6 ): 6923 - 6939 .

GULICH R , KÖHLER M , LUNKENHEIMER P , et al . Dielectric Spectroscopy on Aqueous Electrolytic Solutions [J ] . Radiation and Environmental Biophysics , 2009 , 48 ( 1 ): 107 - 114 .

CHEN T , HEFTER G , BUCHNER R . Dielectric Spectroscopy of Aqueous Solutions of KCl and CsCl [J ] . The Journal of Physical Chemistry A , 2003 , 107 ( 20 ): 4025 - 4031 .

FENG R H , LIN X Y . RFID-based real-time salt concentration monitoring with adaptive EKF [J ] . Sensors , 2025 , 25 ( 12 ): 3826 .

FU S C , LIU W F , ZHANG K , ZHOU Y C . Example-Feature Graph Convolutional Networks for Semi-Supervised Classification [J ] . Neurocomputing , 2021 , 461 : 63 - 76 .

SEMLANI Y , RELAN M , RAMESH K . PCN: A Deep Learning Approach to Jet Tagging Utilizing Novel Graph Construction Methods and Chebyshev Graph Convolutions [J ] . Journal of High Energy Physics , 2024 , 2024( 7 ): 1 - 15 .

VAN BELLE R , VAN DAMME C , TYTGAT H , DE WEERDT J . Inductive Graph Representation Learning for Fraud Detection [J ] . Expert Systems with Applications , 2022 , 193 : 116463 .

吴海宁 . 高盐化工废水处理工艺与参数优化的研究 [D ] . 广州：华南理工大学， 2010 .

WU H N . Research on treatment technology and parameter optimization of high-salinity chemical wastewater [D ] . Guangzhou : South China University of Technology , 2010 .

GENNARELLI G , ROMEO S , SCARFI M R , et al . A microwave resonant sensor for concentration measurements of liquid solutions [J ] . IEEE Sensors Journal , 2013 , 13 ( 5 ): 1857 - 1864 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

暂无数据