Two Layer Markov Model for Prediction of Future Load and End of Discharge Time of Batteries

Abstract

Please refer to published version for the most recent bibliographic citation information. If a published version is known of, the repository item page linked to above, will contain details on accessing it.

State of Power Prediction for Lithium-Ion Batteries in Electric Vehicles via Wavelet-Markov Load Analysis

“…It is worth mentioning that the present study is an extension of the authors' previous research on the prediction of battery end of discharge time [27] based on Markov models. Improving battery electrical-thermal model for accurate representations in different working points, connecting battery model to vehicle powertrain model, designing a wavelet-Gaussian based model parameterisation method for load prediction and experimental validation tests under realistic scenarios are the main contributions of this research compared to the previous one.…”

“…There exists different load prediction approaches for batteries in the literature. Methods based on moving average of the historical data [25], wavelet analysis [26], Markov models [27,28,29] and neural networks [5,30] are examples for characterising the future load. However only in a subset of researches [5,26], these methods have been accompanied with a power prediction mechanism to evaluate their performance.…”

“…Furthermore, experimental validations under different loading and ambient conditions are conducted for Liion 5Ah cylindrical cells. Discussions on the method details and comparisons with the existing methods for power prediction [26,27] are also given.…”

“…Furthermore the method is advantageous over the popular moving average methods [27] because: 1) Moving average methods are lagged regarding the original load and are unable to identify the new changes in trend and regimes. 2) Moving average methods are similar to low pass filters with the main drawback of hindering the transient and intense movements that may turn to be quite important.…”

“…9) Comparisons: To analyze the effect of the load prediction mechanism for the SoAP, the proposed method is compared with two similar cases. The first case is where the load is obtained via wavelet analysis proposed in [26], and the second case is where the future conditions are predicated via Markov model [27]. The results are given in Table 4.…”