Improved Unscented Kalman Filter Based Estimation Of State Of Operating For Lithium Battery

With the continuous development of energy storage technology and the increasing demand for energy storage systems,ensuring the safety,stability,and efficient operation of energy storage systems has become a focal point of research in the industry.Accurate and efficient assessment of the operational state of lithium-ion batteries is a core issue in battery safety management systems.This paper focuses on the estimation of state of charge and health methods for lithium-ion batteries.Firstly,to address the issue of inaccurate representation of battery degradation in capacity health indicators,five battery-related features are utilized,and a new health indicator is established by integrating them using the fruit fly optimization algorithm.In order to tackle the significant dependency on state initialization in traditional Unscented Kalman Filtering,this paper proposes an improved UKF method based on adaptive Levy flight and dual sampling.Specifically,this method generates a larger set of sigma points through two rounds of Unscented Transform,thus resolving the slow convergence issue.The adaptive Levy flight algorithm is employed to optimize the reconstructed sigma points,enabling more accurate state vector and covariance updates.Additionally,the use of the Fourier model reduces computational complexity and simplifies the structure.Secondly,considering the trade-off between time cost and prediction accuracy of circuit models,the Davie-Menton circuit model is selected for modeling.In order to address the problem of identification bias in traditional recursive least squares(RLS)method leading to reduced accuracy,the recursive constrained total least squares method is employed for parameter identification,resulting in more accurate model parameters.To tackle the issue of slow convergence of the state vector during state-of-charge(SOC)estimation,an improved Unscented Kalman Filter(UKF)proposed in this paper is utilized for SOC estimation,aiming to enhance the accuracy and convergence speed of SOC estimation for lithium-ion batteries.Finally,this paper uses the NASA battery dataset for state of health estimation experiments and the battery experimental data of a new energy technology company for state of charge estimation experiments to verify the proposed methods.The results show that the proposed method can overcome the influence of process noise and observation noise covariance,achieve faster convergence and higher filtering accuracy compared with other filtering algorithms.