This content is not included in your SAE MOBILUS subscription, or you are not logged in.

Robust xEV Battery State-of-Charge Estimator Design Using a Feedforward Deep Neural Network

Journal Article
2020-01-1181
ISSN: 2641-9645, e-ISSN: 2641-9645
Published April 14, 2020 by SAE International in United States
Robust xEV Battery State-of-Charge Estimator Design Using a Feedforward Deep Neural Network
Sector:
Citation: Vidal, C., Kollmeyer, P., Naguib, M., Malysz, P. et al., "Robust xEV Battery State-of-Charge Estimator Design Using a Feedforward Deep Neural Network," SAE Int. J. Adv. & Curr. Prac. in Mobility 2(5):2872-2880, 2020, https://doi.org/10.4271/2020-01-1181.
Language: English

Abstract:

Battery state-of-charge (SOC) is critical information for the vehicle energy management system and must be accurately estimated to ensure reliable and affordable electrified vehicles (xEV). However, due to the nonlinear temperature, health, and SOC dependent behaviour of Li-ion batteries, SOC estimation is still a significant automotive engineering challenge. Traditional approaches to this problem, such as electrochemical models, usually require precise parameters and knowledge from the battery composition as well as its physical response. In contrast, neural networks are a data-driven approach that requires minimal knowledge of the battery or its nonlinear behaviour. The objective of this work is to present the design process of an SOC estimator using a deep feedforward neural network (FNN) approach. The method includes a description of data acquisition, data preparation, development of an FNN, FNN tuning, and robust validation of the FNN to sensor noise. To develop a robust estimator, the FNN was exposed, during training, to datasets with errors intentionally added to the data, e.g. adding cell voltage variation of ±4mV, cell current variation of ±110mA, and temperature variation of ±5ºC. The error values were chosen to be similar to the noise and error obtained from real sensors used in commercially available xEVs. The robust FNN trained from two Li-ion cells datasets, one for a nickel manganese cobalt oxide (NMC) cell and the second for a nickel cobalt aluminum oxide (NCA) chemistry cell, is shown to overcome the added errors and obtain a SOC estimation accuracy of 1% root mean squared error (RMSE).