Real-Time State of Charge Estimation Algorithm for Electrical Batteries

While 21st-century society is shifting towards eco-friendly infrastructures, the main
interest of the automotive industry has been transferring to electric vehicles. This
transformation is dependent on the development of reliable battery management systems
(BMSs). A BMS in electric cars provides vital information about the battery
states including the state of charge (SoC). Optimal and robust SoC estimation algorithms
to deploy with minimal effort are vital for the future electric car industry.
The equivalent circuit model (ECM) based SoC estimation algorithms are widely used
in practice. These algorithms suffer from two dominant error sources, i.e., inaccurate
SoC-OCV relationship and input current measurement noise. In the ECM-based SoC
estimation, these error sources have not been fully mitigated. Firstly, we present a
novel technique to construct the SoC-OCV relation, which is eventually converted to a
single parameter estimation problem. The Kalman filter is implemented to estimate the
SoC and the related battery states using the proposed parameter estimation and the
SoC-OCV construction technique. Secondly, we develop a novel technique to mitigate
the error in the current input measurement. The error is calculated based on difference
between the calculated output and the measured output. Correcting the current input
measurement significantly reduces the SoC estimation error.
We validate the proposed algorithms through computer simulations and battery experiments.
The numerical simulations and the battery experiment demonstrate that
the SoC-OCV relationship is accurately constructed in real-time. The SoC estimation
error remains below 2% in numerical simulations whereas the SoC estimation error
remains within 2.5% in the battery experiment. The current noise mitigation algorithm
reduces the SoC error from 11.3% to 0.56% in the numerical simulations. In the
battery experiment, the SoC error is reduced from 1.74% to 1.12%.