Low computational cost method for online parameter identification of Li-ion battery in battery management systems using matrix condition number

Monitoring the state of health for Li-ion batteries is crucial in the battery management system (BMS), which helps end-users use batteries efficiently and safely. Battery state of health can be monitored by identifying parameters of battery models using various algorithms. Due to the low computation power of BMS and time-varying parameters, it is very important to develop an online algorithm with low computational cost. Among various methods, Equivalent circuit model (ECM) -based recursive least squares (RLS) parameter identification is well suited for such difficult BMS environments. However, one well-known critical problem of RLS is that it is very likely to be numerically unstable unless the measured inputs make enough excitation of the battery models. In this work, A new version of RLS, which is called condition memory recursive least squares (CMRLS) is developed for the Li-ion battery parameter identification to solve such problems and to take advantage of RLS at the same time by varying forgetting factor according to condition numbers. In CMRLS, exact condition numbers are monitored with simple computations using recursive relations between RLS variables. The performance of CMRLS is compared with the original RLS through Li-ion battery simulations. It is shown that CMRLS identifies Li-ion battery parameters about 100 times accurately than RLS in terms of mean absolute error.