State of Health is a generalised indicator to establish the ageing conditions of a battery and it is usually calculated by starting from the remaining capacity. Unfortunately, under normal operating conditions, the residual capacity is not measurable, and therefore it is necessary to estimate it. Various techniques have been employed to perform the estimation, both from the model-based and the data-driven domains, with results that are strongly dependent on the underlying model accuracy or on the quality and on the processing of the input predictor vector. In this paper, the capacity estimation problem is tackled by employing ensemble learning techniques with data obtained under operating conditions that are deliberately challenging to perform a data-driven estimate. A comparison framework was settled and the performance of two ensemble learning techniques, called bagging and boosting, were investigated when trained with two different predictor vectors. Analysing and crossing the obtained results, it is possible to show that, while both techniques seem to be insensitive to some peculiar predictor vectors, bagging provides a lower estimation error than the boosting technique. © 2020 IEEE.

A Comparison of Ensemble Machine Learning Techniques for the Estimate of Residual Capacity of Li-Ion Batteries

Guarino, A.;Zamboni, W.;
2020-01-01

Abstract

State of Health is a generalised indicator to establish the ageing conditions of a battery and it is usually calculated by starting from the remaining capacity. Unfortunately, under normal operating conditions, the residual capacity is not measurable, and therefore it is necessary to estimate it. Various techniques have been employed to perform the estimation, both from the model-based and the data-driven domains, with results that are strongly dependent on the underlying model accuracy or on the quality and on the processing of the input predictor vector. In this paper, the capacity estimation problem is tackled by employing ensemble learning techniques with data obtained under operating conditions that are deliberately challenging to perform a data-driven estimate. A comparison framework was settled and the performance of two ensemble learning techniques, called bagging and boosting, were investigated when trained with two different predictor vectors. Analysing and crossing the obtained results, it is possible to show that, while both techniques seem to be insensitive to some peculiar predictor vectors, bagging provides a lower estimation error than the boosting technique. © 2020 IEEE.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11386/4755674
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 3
  • ???jsp.display-item.citation.isi??? 1
social impact