Advanced combustion balance analysis in circulating fluidized bed boilers: A machine learning approach with integrated key performance indicators
No Thumbnail Available
Journal Title
Journal ISSN
Volume Title
Kemian tekniikan korkeakoulu |
Master's thesis
Author
Date
2024-06-10
Department
Major/Subject
Industrial Energy Solutions and Sustainability
Mcode
CHEM3044
Degree programme
Master’s Programme in Advanced Energy solutions (AAE)
Language
en
Pages
71+4
Series
Abstract
This thesis investigates the use of machine learning (ML) techniques for analyzing Circulating Fluidized Bed (CFB) boiler data using Key Performance Indicators (KPIs). To gain insights into the operational dynamics and combustion characteristics of CFB boilers, various methodologies are used, including Cluster Analysis, Self-Organizing Map (SOM) Analysis, Random Forest Method, and Support Vector Machine (SVM) Method. Significant findings about data distribution, relationships between KPIs and underlying data structures, and classification algorithm performance are discovered through analysis of selected KPIs. The Random Forest Method performed exceptionally well, as evidenced by the results, which show 100% accuracy in classifying instances in all cases. The SVM Method with the Radial Basis Function (RBF) kernel, on the other hand, stands out as the best option because of its ability to reliably capture intricate patterns and adapt well to new data. A useful tool for evaluating the combustion balance inside the boiler system is the Combustion Index (CI) metric, was developed. Moreover, validation using new data confirms the reliability of the SVM model, showcasing its ability to accurately predict previously unseen industrial CFB boiler data. Overall, this thesis advances ML methods for the analysis of CFB boiler data, which has implications for improving sustainability and operational effectiveness in industrial settings.Description
Supervisor
Järvinen, MikaThesis advisor
Liukkonen, MikaKeywords
machine learning, circulating fluidized bed boiler, classification, cluster analysis, combustion balance, key performance indicators
