1 / 8

Preliminary Report on Data from CEPSA Tenerife Refinery

This report presents preliminary findings from a data analysis conducted on the CEPSA Tenerife Refinery using advanced prediction methodologies. The study employs an Open Prediction System (OPS) to predict gas consumption and classify various operational events, including heating system anomalies and pump faults. Utilizing techniques such as singular value analysis, neural networks, and support vector machines, the research analyzes training and testing data to derive insights. Initial results indicate promise in applying SVM and NN for improved outcomes, suggesting future enhancements can be made by incorporating additional factors.

vin
Download Presentation

Preliminary Report on Data from CEPSA Tenerife Refinery

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Preliminary Report on Data from CEPSA Tenerife Refinery Olga Štěpánková, Jiří Kléma, Lenka Lhotská step@labe.felk.cvut.cz Gerstner laboratory Czech Technical University in Prague http://cyber.felk.cvut.cz

  2. Overview • introduction • data & methods • results • conclusion

  3. Introduction - OPS Open Prediction System (ops.certicon.cz) • motivation for development • prediction of gas consumption (TDE, Germany) • classification of events in the heating system (Grundfos, Denmark) • classification of faults in pumps (Rockwell Automation, USA) • classification of heart failures (Vitatron Medical, The Netherlands)

  4. Data & Methods • all data (items 9 - 337) were divided into • training data: items 9 - 249 • testing data: items 250 - 337 • OPS (Open Prediction System) applied • singular value analysis (improved regression based method) • neural networks (backpropagation) - 3 layers • support vector machines using sequential minimal optimization (specialized iterative solver) - linear kernel decomposition

  5. Relative error for testing data

  6. Statistics of the achieved results Testing data = 86 Training data = 243

  7. Conclusion • very preliminary results from SVM are promising • combination of SVM and NN could bring good results recommendations for future improvements - backgroud knowledge involved, e.g. • consideration of time factor (reverse pivoting) • consideration of inertia of the system (integration of some attributes) - sliding window

  8. Contacts Czech Technical University in Prague Gerstner Laboratory Jiří Kléma - klema@labe.felk.cvut.cz Olga Štěpánková - step@labe.felk.cvut.cz Lenka Lhotská - lhotska@fel.cvut.cz

More Related