锂电池荷电状态、健康状态以及功率状态的联合在线估计算法

发布时间：2018-04-05 19:32

本文选题：荷电状态　切入点：健康状态　出处：《西南交通大学》2017年硕士论文

【摘要】：锂电池在交通工具、电网、移动设备中被大量应用。然而在汽车领域,由于汽车使用环境差异极大,因此,锂电池作为它的一种主要动力来源,不但需要在一些复杂的外部条件,而且需要在这些苛刻的使用条件下甚至是在遭受破坏时,保证驾驶者以及乘客的安全。在这一过程中,要保证电池的安全使用,并提供更好的电池管理策略,需要对电池自身状态进行估计。常见锂电池状态描述变量有荷电状态(State of Charge,SoC)、电池的健康状态(State of Health,SoH)以及电池的功率状态(State of Power,SoP)。三者分别描述电池可持续对外供电或受电的能力、电池剩余使用寿命以及电池即时对外进行输出电能或接受电能的能力。虽然目前对于这三个状态量的在线估计有大量的估计算法,但是汽车内部嵌入式系统有限的计算能力限制了对诸多算法的使用。针对这一问题,给出一种联合性算法,在统一使用改进后的Randle电池模型的基础上,对三个状态量进行在线估计。对于电池模型,一种标准化的迭代最小二乘算法被用于对电池模型参数进行识别并避免了最小二乘算法中,由于协方差矩阵反复迭代而导致的数据溢出问题。而识别的电池参数则被用于直接估计电池的健康状态。由于实际情况中,噪声的功率以及其对应的协方差矩阵是难以获知的,因此,一种带有自适应噪声协方差矩阵的拓展卡尔曼滤波(Extended Kalman Filter,EKF)算法被用于电池进一步的SoC在线估计。估计过程中,电池开路电压(Open Circuit Volt-age,OCV)也被作为状态变量归并入迭代过程中。其次,结合电池模型以及改进后EKF算法中的状态方程,一种基于多限制条件下峰值功率的估计算法被给出。其计算过程涵盖对电压电流以及电池SoC处于极限状态下的峰值功率计算。其结果结合电池额定功率被最终用于SoP的估计。三个部分的算法的准确性均有仿真实验加以验证。最后,硬件实现的电池数据采集卡配合BTS-5V300A设备通过实验对电池模型参数识别算法以及电池SoC、SoH与SoP估计算法的估计效果进行了验证。仿真实验与电池实验结果的对比,验证了估计算法的有效性。
[Abstract]:Lithium batteries are widely used in vehicles, power grids and mobile devices.However, in the field of automobile, because of the great difference in the environment of automobile use, lithium battery is one of its main power sources, not only in some complicated external conditions, but also in some complicated external conditions.It is also necessary to ensure the safety of drivers and passengers in these harsh conditions, even in the event of destruction.In order to ensure the safe use of the battery and provide a better battery management strategy, it is necessary to estimate the state of the battery itself.The commonly described state variables of lithium batteries are the charged state of Chargeof SoC, the healthy state of the battery and the power state of the battery.They describe the ability of the battery to supply or receive electricity from the outside, the remaining life of the battery and the ability of the battery to output or receive the electric energy immediately.Although there are a large number of estimation algorithms for these three state variables online, the limited computing power of the embedded system limits the use of many algorithms.To solve this problem, a joint algorithm is proposed to estimate the three state variables on line on the basis of unified use of the improved Randle battery model.For the battery model, a standardized iterative least squares algorithm is used to identify the parameters of the battery model and to avoid the problem of data overflow caused by repeated iteration of the covariance matrix in the least squares algorithm.The identified battery parameters are used to estimate the health status of the battery directly.Because the power of noise and its corresponding covariance matrix are difficult to obtain in practice, an extended Kalman filter extended Kalman filter (EKF) algorithm with adaptive noise covariance matrix is used for further SoC on-line estimation of batteries.During the estimation process, the open Circuit voltage of the cell is also incorporated into the iterative process as a state variable.Secondly, combining the battery model and the state equation of the improved EKF algorithm, a new algorithm for estimating the peak power based on multiple constraints is presented.The calculation process includes the calculation of voltage and current and the peak power of SoC in the limit state.The results combined with battery rated power are ultimately used to estimate SoP.The accuracy of the algorithm is verified by simulation experiments.Finally, the hardware implementation of the battery data acquisition card and BTS-5V300A equipment through experiments to identify the parameters of the battery model algorithm and the battery SoH and SoP estimation algorithm is verified.The effectiveness of the estimation algorithm is verified by comparing the simulation results with the battery experiments.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM912

【参考文献】