A novel interval‐based approach for quantifying practical parameter identifiability of a lithium‐ion battery model

Abstract

Summary Practical identifiability of battery model parameters, on which both modeling accuracy and robustness rely, is considered as a very important prerequisite for advanced onboard monitoring and control of Lithium‐ion batteries. In this paper, a novel confidence‐interval‐based approach is proposed for the quantification and assessment of the practical identifiability of a widely used second order battery equivalent circuit model (ECM). This method utilizes profile likelihood and likelihood ratio subset statistic to calculate each parameter's confidence interval, based on which a normalized index is further derived for facilitating quantification and fast comparison of the identifiability degree among different parameters. Using this approach, the practical identifiability of the second order ECM under lab‐collected experimental data is successfully evaluated, and the influences of several real‐world factors are systematically examined through extensive simulations. The results show that the open circuit voltage and ohmic internal resistance have a much larger degree of identifiability in all the investigated conditions. Some practically useful insights on performing battery parameter identification are also provided.

A novel feedback correction-adaptive Kalman filtering method for the whole-life-cycle state of charge and closed-circuit voltage prediction of lithium-ion batteries based on the second-order electrical equivalent circuit model

“…In Equation (7), 𝑡 and 𝑡 − 1 are the dependent current and previous time points, respectively. Combining the parameters of CCV and current, the iterative calculation equation is obtained, as shown in Equation (8).…”

“…The supervised machine learning model is constructed based on the second-order ECM (SO-ECM) to achieve an accurate fault diagnosis for batteries of electric aircraft under different test conditions [7]. A confidence-interval-based prognostic model is proposed to quantify and assess the practical parameter identification ability of the SO-ECM using profile likelihood and ratio subset statistics [8]. A fractional-order ECM is established for multiple lithium-ion batteries under different states based on electrochemical impedance spectroscopy (EIS) analysis [9].…”

Improved covariance matching—electrical equivalent modeling for accurate internal state characterization of packing lithium‐ion batteries

“…Consequently, the nonlinearity degree returns to be normal with high accuracy. Therefore, the global convergence is calculated and stretched, in which the state-space equations are used relatively and the Kalman gain matrix is obtained, as shown in Equation (26).…”

“…24,25 To simulate the responding voltage characteristics under different power supply conditions, the equivalent modeling is divided into black box, electrochemical, and electrical circuit types. The black box modeling is a kind of nonlinear treatment to describe the voltage-response characteristics, [26][27][28] which includes neural networks (NN), support vector machines (SVM), and so on. The black box model is trained by the real-time measured data, depending on the experimental test seriously.…”

On structural and practical identifiability

“…Moreover, it is especially crucial to ensure that model predictions are well-determined. It is analyzed increasingly often to judge a model's predictivity [40,41,42,43,44]. The notion of practical identifiability has been rather vague in the literature, mainly referring to large confidence intervals [45,46,47].…”