Optimized calibration of P2D lithium ion battery physical model following sensitivity-based multi-measurement protocol

Summary:

This poster deals with the activities ongoing at Politecnico di Milano in the framework of the European Project DigiPrime, on lithium-ion batteries diagnostics and estimation of residual capabilities at end-of-life. It presents the development of a fast multi-measurement protocol that can ease the calibration of the pseudo-two-dimensional model (P2D) with a thermal model, to enable the simulation of battery operation in broad combinations of operating conditions.
The first part involves a one-factor-at-time sensitivity analysis. The model response due to a change of 28 parameters (geometric and thermodynamic excluded) is studied at different operating conditions and for three techniques: discharges, relaxation after discharge and electrochemical impedance spectroscopy (EIS). It highlights the conditions of maximum and minimum sensitivity of many parameters and enables to exclude from the calibration 14 of them, which can be taken equal to a literature value without affecting the calibration.
As a result, a multi-measurement protocol is developed, including 2 EIS, 2 discharges and one relaxation at 2 temperature levels, within 2h of overall test. It is applied to two pristine samples, a high-power 26650 2.6Ah NMC+LMO (case A) and a high-energy 18650 2.25Ah NMC sample (case B). For the latter the protocol is slightly changed due to the lower maximum discharge C-rate. Experimental data are fed as input to a Particle Swarm Algorithm, which, starting from random values, progressively approach a solution that minimizes the difference between model and experimental data. The calibration includes three steps: first, kinetic constants and double layer capacitances are calibrated on the kinetic loop of the impedance spectra, then solid diffusivities and electrolyte conductivity on discharges and relaxation. Lastly, a little refinement is required on kinetic constants due to the conductivity value obtained at the second step.
Solution repeatability is achieved and its reliability is proved by way of a wide validation dataset. The model reproduces with great accuracy a broad set of full discharges, EIS and a dynamic profile. To further highlight the strength of this method, another calibration is performed employing only 5 full discharges as input data to the algorithm, as representative of a literature approach. The result shows that the calibrated model reproduces experimental discharges with comparable accuracy, but it fails with EIS. Analysing the parameters values and its repeatability, the discharge-based calibration can reliably identify the values of diffusive properties only, because, as highlighted by the sensitivity analysis, the identifiability of kinetic and resistive properties is not so high along the discharge, but it is the highest on the impedance spectra.
Therefore, these successful results deserve to be employed on aged cells, to provide support in the ageing interpretation through a reliably calibrated physical model.