The Comparison of SOC Estimation Performance with Different Input Parameters Using Neural Network

Author(s): 
X. Zhao, S. Wang, M. Yu

Source:

X. Zhao, S. Wang, & M. Yu, “The Comparison of SOC Estimation Performance with Different Input Parameters Using Neural Network”, Journal of Mechanical Engineering Research and Developments, vol. 39, no. 3, pp. 730-735, 2016. 
 

ABSTRACT: The battery is a high nonlinear system and the neural network model (NNM) has been proposed to estimate the battery’s State-Of-Charge (SOC) recently. To determine different input parameters’ impact on NNM’s estimation performance, the paper firstly identifies the battery’s external parameters affecting the value of SOC and puts forwards three different NNMs for estimation performance comparison. And then a variety of discharging processes are experimented on 6Ah Lithium-ion battery to collect the training and testing data samples. Lastly, based on the data samples, the SOC estimations using NNM are conducted and estimation performances for three models are contrasted. The results show that the battery’s temperature and internal resistance both could improve NNM’s estimation precision and robustness against the measuring noise. But the temperature is more suited to estimate SOC in practice for its easy implement in practice.

Keywords :  Electric vehicle; State-of-charge; Neural network; Measuring noise; Robustness.