A novel online model parameters identification method with anti‐interference characteristics for lithium‐ion batteries

Abstract

Model-based state of charge (SOC) estimation method depends on the accuracy of the online identified battery model. However, when battery model parameters are identified by conventional recursive least squares (RLS), voltage and current noise will lead to the deviation of parameters and further affect the accuracy of SOC estimation. To analyze the difference influence of voltage and current noise, a simulation is carried out under the same noise variance of both voltage and current. Results indicate that voltage noise is the main cause of parameter deviation. Based on these results, a novel method, the auxiliary model recursive least square (AM-RLS) which enjoys low computational complexity is proposed to compensate the parameter deviation caused by voltage noise. The AM-RLS method is further combined with an extendedKalman filter (EKF) to estimate SOC in real time. Both simulation and experimental results show that AM-RLS can effectively attenuate the parameter deviation and SOC estimation error under noise interference. Compared with RLS, the maximum SOC errors of the proposed method are reduced by 2.6% and 5.86% respectively under two current profiles.

A novel streamlined particle‐unscented Kalman filtering method for the available energy prediction of lithium‐ion batteries considering the time‐varying temperature‐current influence

“…The realization process mainly adopts the Bayesian criterion to weigh discrete random samples and complete the mean integral operation. The normalized weight is calculated according to Equation (26).…”

“…The constrained Bayesian Double Filtering method has a good effect on the available energy prediction of lithium-ion batteries effectively. 26 The prediction can be realized effectively by the embedded Square Root-Volume Kalman Filtering algorithm, which is adaptive to the time-varying power supply conditions of all climate batteries. 27 Intelligent Wavelet Neural Network and electro-thermal modeling strategies are constructed to achieve state prediction based on discrete wavelet transform.…”

“…Using the backtracking search algorithm, the NN model can be established to express the energy state difference of the internal series cells, and the observation ability can be improved in exchange for shorter calculation time and accuracy. The constrained Bayesian Double Filtering method has a good effect on the available energy prediction of lithium‐ion batteries effectively 26 . The prediction can be realized effectively by the embedded Square Root‐Volume Kalman Filtering algorithm, which is adaptive to the time‐varying power supply conditions of all climate batteries 27 .…”

Optimal parameter identification strategy applied to lithium‐ion battery model

“…A state‐space battery model including frequency domain characteristics was used to identify the battery parameters in Reference 13, although in its performance, this computational method is very complex. An online parameter identification method with anti‐interference characteristics that eliminates the voltage and current noise was presented in Reference 21. Nevertheless, real‐time identification is vulnerable to identification error during the identification process.…”

A novel intelligent weight decreasing firefly–particle filtering method for accurate state‐of‐charge estimation of lithium‐ion batteries

“…19 There are three kinds of commonly used lithiumion battery models: electrochemical model, data-driven model, and equivalent circuit model. 20 Electrochemical model is often used to describe the electrochemical reaction process inside the battery, 21 which can deeply reflect the complex electrochemical kinetic mechanism. 22 However, it is too complex, large amount of calculation, and difficult to solve online.…”