一种基于自适应迁移DeCPLS模型的变工况化工过程软测量方法
By using the adaptive transfer DeCPLS model, the problems of low prediction accuracy, high negative transfer risk, and lack of online adaptation in chemical processes are solved. This method enables efficient cross-condition data fusion and online adaptive correction for chemical processes, thereby improving the model's generalization ability and prediction accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA UNIV OF PETROLEUM (EAST CHINA)
- Filing Date
- 2026-05-19
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies lack cross-condition migration capabilities in chemical processes, pose a risk of negative migration, lack online adaptive capabilities, and are difficult to align heterogeneous data, resulting in low prediction accuracy and degraded model performance.
A method based on the adaptive transfer DeCPLS model is adopted to achieve effective fusion and online adaptive correction of cross-condition data through clustering of historical operating condition data, distribution difference measurement, similarity domain selection, feature alignment and weighted ensemble modeling.
It significantly improves the accuracy of cross-condition prediction for chemical processes, suppresses the risk of negative migration, maintains the local geometric structure of the data, achieves lightweight online adaptiveness, and enhances the generalization performance of the model.
Smart Images
Figure CN122220907B_ABST
Abstract
Citation Information
Patent Citations
Multi-working-condition process soft measurement modeling method based on feature migration learning
CN109060001A
Complex industrial process soft measurement modeling method based on transfer learning random configuration network
CN118194557A