Please use this identifier to cite or link to this item: http://theses.ncl.ac.uk/jspui/handle/10443/5630
Full metadata record
DC FieldValueLanguage
dc.contributor.authorXu, Jin-
dc.date.accessioned2022-12-02T15:36:59Z-
dc.date.available2022-12-02T15:36:59Z-
dc.date.issued2022-
dc.identifier.urihttp://hdl.handle.net/10443/5630-
dc.descriptionPh. D. Thesis.en_US
dc.description.abstractThere have been many recursive algorithms investigated and introduced in real time parameter estimation of Switch Mode Power Converters (SMPCs) to improve estimation performance in terms of faster convergence speed, lower computational cost and higher estimation accuracy. These algorithms, including Dichotomous Coordinate Descent (DCD) - Recursive Least Square (RLS), Kalman Filter (KF) and Fast Affine Projection (FAP), etc., are commonly applied for performance comparison of system identification of single-rail power converters. When they need to be used in multi-rail architectures with a single centralized controller, the computational burden on the processor becomes significant. Typically, the computational effort is directly proportional to the number of converters/rails. This thesis presents an iterative decimation approach to significantly alleviate the computational burden of centralized controllers applying real-time recursive system identification algorithms in multirail power converters. The proposed approach uses a flexible and adjustable update rate rather than a fixed rate, as opposed to conventional adaptive filters. In addition, the step size/forgetting factors are varied, as well, corresponding to different iteration stages. As a result, reduced computational burden and faster model update can be achieved. Recursive algorithms, such as Recursive Least Square (RLS), Affine Projection (AP) and Kalman Filter (KF), contain two important updates per iteration cycle. Covariance Matrix Approximation (CMA) update and the Gradient Vector (GV) update. Usually, the computational effort of updating Covariance Matrix Approximation (CMA) requires greater computational effort than that of updating Gradient Vector (GV). Therefore, in circumstances where the sampled data in the regressor does not experience significant fluctuations, re-using the Covariance Matrix Approximation (CMA), calculated from the last iteration cycle for the current update can result in computational cost savings for real- time system identification. In this thesis, both iteration rate adjustment and Covariance Matrix Approximation (CMA) re-cycling are combined and applied to simultaneously identify the power converter model in a three-rail power conversion architecture. Besides, in multi-rail architectures, due to the high likelihood of the at-the-same-time need for real time system identification of more than one rail, it is necessary to prioritize each rail to guarantee rails with higher priority being identified first and avoid jam. In the thesis, a workflow, which comprises sequencing rails and allocating system identification task into selected rails, was proposed. The multi-respect workflow, featured of being dynamic, selectively pre-emptive, cost saving, is able to flexibly change ranks of each rail based on the application importance of rails and the severity of abrupt changes that rails are suffering to optimize waiting time and make-span of rails with higher priorities.en_US
dc.language.isoenen_US
dc.publisherNewcastle Universityen_US
dc.titleOptimization of System Identification for Multi-Rail DC-DC Power Convertersen_US
dc.typeThesisen_US
Appears in Collections:School of Engineering

Files in This Item:
File Description SizeFormat 
Xu Jin e-copy.pdfThesis13.29 MBAdobe PDFView/Open
dspacelicence.pdfLicence43.82 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.