Estimation of SOC for Battery in Electric Vehicle Based on STUKF Algorithm

Abstract

Lithium-ion (Li-on) battery state of charge (SOC) estimation is important for electric vehicles (EVs). To eliminate the effects of colored noise on SOC estimation, a new estimation method that based on Unscented Kalman Filter (UKF) Algorithm is proposed for high-power Li-ion batteries. First of all, based on the battery chemical properties, this paper established the improved PNGV battery model and identified the battery parameters. Then, accuracy of the model was verified under UDDS working condition. Finally, according to the influence of colored noise on estimating SOC of battery by the Unscented Kalman Filter (UKF) Algorithm, this paper proposed the Strong Tracking Unscented Kalman Filter (STUKF) Algorithm and introduced the fading factor. which forces the innovation sequence to be orthogonal and strengthens the correction of the state estimation by the new data. The result of simulation shows the STUKF Algorithm has batter tracking characteristic on estimating SOC of battery. Introduction Along with the continuous intensification of a series of problems such as energy exhaustion, environmental pollution and greenhouse effect, the development of electric vehicles has attracted wide attention of the automobile industry, and various major automobile manufacturers have produced new electric vehicles one after another. As one of the three key technologies of electric vehicles, battery management technology has become the focus of research in major automobile enterprises, universities and research institutes [1]. Battery management system of electric vehicle is mainly responsible for battery status detection, power balance, fault protection, etc. As an important indicator of balance and fault diagnosis, SOC is related to the working stability of the whole battery management system. Therefore, the accuracy of estimating SOC of battery is particularly important [2]. The battery state of charge (SOC) is the most important point of the battery management system (BMS), whose estimation methods are often broken through in models and algorithms [3]. At present, there are many methods of building models for estimating SOC of battery. Paper [4] introduced the Rint model, Thevenin model, PNGV model and GNL model, compared with two other models, Rint model and Thevenin model are more simplified, so that the accuracy of the model was not exact. On the other hand, although the GNL model has high precision, the calculation is too much when it is applied. Besides improving the model, estimation algorithm is also very important. In [5], the SOC and capacity of batteries are estimated by double-observation algorithm under the application of reduced-order electrochemical model of composite-electrode batteries. In recent years, there have been many advanced algorithms have been applied in estimating SOC of battery. such as Extended Kalman Filter (EKF)algorithm, Unscented Kalman Filter (UKF)algorithm and NARX Neural Network algorithm. In [6], The paper proposed non-linear autoregressive control method (NARX) with the exogenous input, which is effective and large computation for the control of the system. In [7], the author used extended Kalman Filtering algorithm to estimate SOC of battery with the PNGV model. In this model, RC circuit was added, however, the capacitance did not participate in identification which describes open circuit voltage variations because of charge accumulation. EKF algorithm depends more strongly on the accuracy of the model parameters, which are difficult to achieve, so that the estimated results are not accurate because the EKF algorithm can only simplify the system into a first order approximation model. Due to the disadvantages of Extended Kalman International Conference on Modeling, Analysis, Simulation Technologies and Applications (MASTA 2019) Copyright © 2019, the Authors. Published by Atlantis Press. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). Advances in Intelligent Systems Research, volume 168