Battery state of charge estimation using adaptive subspace identification method

Estimation of battery state of charge (SOC) is essential for many emerging battery powered applications such as smart phones, electric and hybrid electric vehicles. In this paper, we propose a new battery SOC estimation method using adaptive subspace identification method. The subspace identification method is a numerically robust approach and is used to build the dynamic linear model based on battery´s terminal voltages and current. To deal with the nonlinearity of the battery, the transient battery terminal voltages and current are partitioned into piecewise linear regions and subspace identification is performed on each linear region. As a result, the battery SOC can be accurately calculated for each region. Our experiments show that the new method has an error margin of 1.4% from ideal SOC values as given by Dualfoil, a powerful battery simulator. This outperforms the least square estimation algorithm, which is found to have a higher error margin of 4.5% for some load profiles, while not converging at all for some other load profiles.