State of charge estimation of li-ion batteries based on the noise-adaptive interacting multiple model
This paper presents a type of noise-adaptive (NA) interacting multiple model (IMM) algorithm combined with an unscented Kalman filter (UKF) in order to address problems in poor filtering accuracy and filtering divergence of IMM caused by the statistical properties of noise. These properties further...
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
Elsevier
2021
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/c452c9c3d1bc41ccbd19bdf5ea74b950 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | This paper presents a type of noise-adaptive (NA) interacting multiple model (IMM) algorithm combined with an unscented Kalman filter (UKF) in order to address problems in poor filtering accuracy and filtering divergence of IMM caused by the statistical properties of noise. These properties further affect the estimation accuracy of state of charge (SOC) when IMM deals with dynamic changes in battery model parameters. Accordingly, the proposed algorithm can realize the accurate estimation of SOC when model parameters change dynamically and when the statistical properties of noise are unknown. By integrating a Sage-Husa noise estimator, NA-IMM-UKF enabled the whole UKF model set to estimate and correct noise information in real time in order for posteriori and unknown noise information to be adjusted adaptively. At the same time, a forgetting factor was introduced in order to optimize the proposed algorithm, thus improving the problem in which the Sage–Husa noise estimator converges slowly when used in conjunction with UKF. By conducting an experiment and simulation, NA-IMM-UKF was shown to carry out SOC estimation under multiple models, with an average error of only 0.4% and maximum error of only 1.08%. However, by comparing the estimated result of SOC under a single model with the Sage–Husa estimator minus the forgetting factor, the average error dropped by 0.15% while the maximum error decreased by 2.78%. In the final noise comparison experiment, following the addition of unknown random noise, the average error of the NA-IMM-UKF algorithm was found to be only 0.48%, while the maximum error was only 1.51%, far surpassing the estimation results of the IMM-UKF algorithm in the same state. As a result, even if the statistical properties of noise are uncertain, the proposed algorithm can still estimate SOC both accurately and effectively. |
|---|