Multi-Method Fault Detection of Cooling Fan in Control Cabinet using Temperature Analysis
Authors
Hansen, Lasse Bonde ; Keenan Odena, Michael Joseph
Term
4. semester
Education
Publication year
2023
Submitted on
2023-06-01
Pages
69
Abstract
Denne afhandling udvikler og afprøver algoritmer til fejldetektion og -diagnose i køleventilatorer, der holder temperaturen nede i vindmøllers kontrolskabe. For at efterligne et virkeligt skabs termodynamiske adfærd blev der bygget to simulationsmodeller: en enkeltzone-model og en flerzone-model. Modellerne gengiver temperaturforløb i både skabet og nacellen (gondolen) og leverer temperaturdata. Systemparametre blev identificeret i en kalibreringsfase med Recursive Least Squares (RLS), en metode der løbende tilpasser modelparametre ud fra temperaturmålinger. De estimerede parametre og målinger blev derefter brugt i flere metoder til fejldetektion og -diagnose (FDD): en bank af observatører (flere estimeringsfiltre, der sammenligner forventede og målte temperaturer), Multiple Model Adaptive Estimation (MMAE, der afvejer flere samtidige modeller) og fælles tilstandsestimering (en metode der estimerer flere størrelser samtidigt). Med disse metoder opnåedes pålidelig temperaturestimering og en vis evne til at forudsige køleventilatorernes sundhedstilstand inden for et vist interval.
This thesis develops and tests algorithms to detect and diagnose faults in cooling fans that keep wind turbine control cabinets at safe temperatures. To mimic the thermal behavior of a real cabinet, two simulation models were built: a single-zone model and a multi-zone model. These models reproduce temperature trends in both the cabinet and the nacelle (the housing at the top of the turbine) and generate temperature data. System parameters were identified in a calibration phase using Recursive Least Squares (RLS), a method that continuously tunes model parameters based on temperature measurements. The estimated parameters and measurements were then used in several fault detection and diagnosis (FDD) methods: a bank of observers (multiple estimation filters that compare expected and measured temperatures), Multiple Model Adaptive Estimation (MMAE, which weighs several models in parallel), and joint state estimation (a method that estimates several system quantities together). Using these methods, the study achieved reliable temperature estimation and an ability to predict the cooling fans’ health status within a certain range.
[This summary has been rewritten with the help of AI based on the project's original abstract]
Keywords
Documents