A comparative study and review of different Kalman filters by applying an enhanced validation method. (November 2016)

• Record Type: Journal Article
• Authors: Campestrini, Christian; Heil, Thomas; Kosch, Stephan; Jossen, Andreas
• Is Part Of: Journal of energy storage. Volume 8(2016)
• Page Start: 142
• Page End: 159
• Publication Date: 2016-11
• Subjects: Battery modelling -- Parameter estimation -- Lithium-ion -- State of charge -- Benchmark -- Filter tuning; Energy storage -- Periodicals; Energy storage -- Research -- Periodicals; 621.3126
• DOI: 10.1016/j.est.2016.10.004
• Languages: English
• ISSNs: 2352-152X

Abstract

Abstract : Highlights: Implementation of 18 different Kalman filters. Comparison of these filters with an enhanced validation method. Consideration and evaluation of the most important cell dependencies within the validation. The importance of the filter tuning is presented. Abstract: The Kalman filter is a common state of charge estimation algorithm for lithium-ion cells. Since its first introduction in the application of lithium-ion cells, different implementations of Kalman filters were presented in literature. However, due to non-uniform validation methods and filter tuning parameters, the performance of different Kalman filters is difficult to quantify. On this account, we compare 18 different implementations of Kalman filters with an enhanced validation method developed in our previous work. The algorithms are tested during a low-dynamic, high-dynamic and a long-term current load profile at −10 °C, 0 °C, 10 °C, 25 °C and 40 °C with a fixed set of filter tuning values. To ensure comparability, a quantitative rating technique is used for estimation accuracy, transient behaviour, drift, failure stability, temperature stability and residual charge estimation. The benchmark shows a similar estimation accuracy of all filters with an one and two RC term equivalent circuit model. Furthermore, a strong dependency on temperature during high-dynamic loads is observed. To evaluate the importance of the tuning parameters, the temperature dependency is reduced with an individual filter tuning. It is reasoned, that not only the filter type is significant for the estimation performance, but the filter tuning.