A New Optimization Algorithm for a Li-ion Battery Equivalent Electrical Circuit Identification

An equivalent electrical circuit (EEC) is commonly used to model the electrochemical aspect of a battery. Its structure can be more or less complicated depending on the needs of the application. Impedance measurements obtained by electrochemical impedance spectroscopy need non integer order impedance in order to correctly model passivation film and diffusion phenomena. Identification of the parameters of the EEC from electrochemical spectroscopy is difficult because of non unicity of the solution and lack of convergence of standard algorithms. The study of the experimental measurements shows that each chemical phenomenon is sollicited in a precise waveband. We propose thus three approaches to improve the identification convergence depending on whether / how we introduce information on the frequency or not. In the first one, we identify all the parameters in the whole waveband (classical approach). The second and third ones (alternative approaches) split the waveband and identify only the corresponding excited parameters. In all these cases, the identification of the parameters of the EEC should use a non linear method and must lead to an accurate result. The Levenberg-Marquardt algorithm is adopted in this study. A statistical study shows that the first approach can’t converge strongly while the second and third one converge with the presence of a bias. Thus we propose to adopt a new method of identification based on a two-steps algorithm.