Battery equivalent circuit models (ECMs) are widely employed in online battery management applications. The model parameters are known to vary according to the operating conditions, such as the battery state of charge (SOC). Therefore, online recursive ECM parameter estimation is one means that may help to improve the modelling accuracy. Because a battery system consists of both fast and slow dynamics, the classical least squares (LS) method, that estimates together all the model parameters, is known to suffer from numerical problems and poor accuracy. The aim of this paper is to overcome this problem by proposing a new decoupled weighted recursive least squares (DWRLS) method, which estimates separately the parameters of the battery fast and slow dynamics. Battery SOC estimation is also achieved based on the parameter estimation results. This circumvents an additional full-order observer for SOC estimation, leading to a reduced complexity. An extensive simulation study is conducted to compare the proposed method against the LS technique. Experimental data are collected using a Li ion cell. Finally, both the simulation and experimental results have demonstrated that the proposed DWRLS approach can improve not only the modelling accuracy but also the SOC estimation performance compared with the LS algorithm. •A novel adaptive estimation method of battery ECM parameters and SOC is presented.•The battery parameters of the fast and slow dynamics are estimated separately.•The proposed method does not require a full-order observer for SOC estimation.•The battery modelling and SOC estimation accuracy is improved.•The proposed algorithm is suitable for both offline and online implementation.