金绍琴1 唐莉丽1 吴菁1,2 李明珠2 黄道平3
（1.贵州民族大学数据科学与信息工程学院，贵州 贵阳 550025
2.海口经济学院腾竞依智网络学院，海南 海口 570203
3.华南理工大学自动化科学与工程学院，广东 广州 510640)
摘要：针对污水处理厂的过程数据时变性较大、非线性较强，传统的离线模型难以应对实际处理过程中的工况变化等问题，提出一种基于时差-即时学习的相关向量机（RVM）模型TD-JIT-RVM。通过时间差分（TD）建模提取过程变量之间的关联关系，采用即时学习（JIT）解决时滞引起的模型退化问题。利用TD-JIT-RVM模型对仿真数据集和真实的工业数据集进行分析，结果表明，该模型在两个数据集中比RVM基础模型的RMSE分别提高了94.59%和82.26%。
关键词：污水处理；时间差分；即时学习；相关向量机；在线软测量
中图分类号：TP 277 文献标志码：A 文章编号：1674-2605(2023)05-0004-10
DOI：10.3969/j.issn.1674-2605.2023.05.004
Research on Soft Sensor Modeling of Relevance Vector Machine Based on Time Difference-Just in Time
JING Shaoqin1 TANG Lili1 WU Jing1,2 LI Mingzhu2 HUANG Daoping3
(1.College of Data Science and Information Engineering, Guizhou Minzu University, Guiyang 550025, China 2.TJ-YZ School of Network Science, Haikou University of Economics, Haikou 570203, China
3.School of Automation Science and Engineering, South China University of Technology, Guangzhou 510640, China)
Abstract: A relevance vector machine (RVM) model TD-JIT-RVM based on Time Difference-Just in Time is proposed to address the issues of large time-varying and strong nonlinearity of process data in sewage treatment plants, and traditional offline models are difficult to cope with changes in actual processing conditions. Extract the correlation between process variables through Time Difference (TD) modeling, and solve the model degradation problem caused by time delay using just in time (JIT). The TD-JIT-RVM model was used to analyze the simulation dataset and the real industrial dataset, and the results showed that the RMSE of the model increased by 94.59% and 82.26% compared to the RVM basic model in both datasets, respectively.
Keywords: wastewater treatment; time difference; just in time; relevance vector machine; online soft sensors
