【作者】 张雨；

【导师】 宋春宁； 周晓华；

【作者基本信息】 广西大学 ， 电子信息（专业学位）， 2023， 硕士

【摘要】 随着“碳达峰,碳中和”概念的提出,储能电池领域得到了很大程度地重视。单液流锌镍电池作为储能新技术,由于其拥有高效性、长寿命、可扩展性等优点,这种电池被认为是一种有前途的电化学储能技术,具有广泛的应用前景。电池管理系统(BMS)是电池研究的核心组成部分,也是维持电池安全可靠运行的保障。而电池状态估计处于BMS技术的关键环节,其研究具有重要意义。要想实现电池状态的准确估计,一个合适的电池模型是不可缺少的。本文基于这个问题,进行了各种电池建模方案的讨论,综合电池建模的准确度和快速性,最终选用二阶RC等效电路模型来模拟单液流锌镍电池性能特征。使用恒流脉冲实验数据测试该模型,实验结果表明了该模型的有效性。然而当放电倍率发生变化时,原有二阶RC等效电路模型输出结果不再能很好地跟随参考电压。此时,本文提出了自适应模型,通过耦合放电倍率变量到原有的二阶RC等效电路模型,从而得到参数自适应变化的自适应模型。通过恒流脉冲工况与变电流脉冲工况测试两种模型,实验结果表明自适应模型的有效性与准确性。荷电状态(SOC)作为BMS技术的关键变量,其估计结果的准确性会很大程度地影响电池的其它管理。然而SOC是单液流锌镍电池的内部状态变量,传感器无法直接获得变量数值,因此本文探讨了单液流锌镍电池的多种SOC估计方法,最后选择了基于奇异值分解的无迹卡尔曼滤波算法(SUKF)。该算法通过引入奇异值分解(SVD)方法到UKF算法的协方差更新过程当中,解决了传统无迹卡尔曼滤波算法(UKF)状态协方差矩阵P易陷于非正定、非对称状态而导致求解失败的问题。考虑到单独的SUKF算法估计SOC忽略了电池模型参数的时变性,因此本文提出使用SUKF联合其它滤波算法,组成双层滤波估计算法(如SUKF-EKF、SUKF-FRLS、DSUKF),在估计SOC的同时,也可以估计电池的模型参数。通过恒流脉冲工况与变电流连续工况测试联合算法的有效性,实验结果证明联合估计算法无论在平均绝对误差(MAE)还是在均方根误差(RMSE)方面,都要远低于SUKF估计算法。由于卡尔曼滤波算法的估算性能在很大程度上受噪声及状态协方差矩阵变量初值的影响,然而传统方法对矩阵变量初值的选择是通过经验选定,不能最大程度发挥DSUKF算法的估算性能。因此,本文使用近些年广泛使用的哈里斯鹰优化算法(HHO)优化初始协方差矩阵参数,并引入多策略结合的方法,提出了改进的哈里斯鹰优化算法(IHHO),进一步加强了算法的寻优能力。与此同时,本文将容量更新公式引进到SOC估计算法当中,生成了改进的双层SVD无迹卡尔曼滤波算法(IDSUKF),新的估计算法不仅提高了SOC估计精度,而且还对老化状态(SOH)进行了预测,实现了SOC和SOH的联合估计。更多还原

【Abstract】 With the concept of "carbon peaking and carbon neutral",the field of energy storage batteries has received a great deal of attention.Single-flow zinc-nickel battery is a new technology for energy storage,which is considered as a promising electrochemical energy storage technology with wide application prospects due to its high efficiency,long life and scalability.Battery management system(BMS)is a core component of battery research,is also the guarantee to maintain the safe and reliable operation of the battery.And battery state estimation is in the key link of BMS technology,and its research has great significance.A suitable battery model is indispensable to achieve an accurate estimation of the battery state.Based on this problem,various battery modeling schemes are discussed in this paper,and the second-order RC model is finally selected to simulate the performance characteristics of single-flow zinc-nickel batteries by integrating the accuracy and rapidity of battery modeling.The model was tested using constant current pulse experimental data,and the experimental results showed the validity of the model.However,when the discharge multiplier changes,the output of the original second-order RC model no longer follows the reference voltage well.At this point,this thesis proposes an adaptive model by coupling the discharge multiplier variable to the original second-order RC model,thus obtaining an adaptive model with adaptive parameter changes.Both models are tested by constant current pulse condition and variable current pulse condition,and the experimental results show the effectiveness and accuracy of the adaptive model.SOC as a key variable in BMS technology,the accuracy of the estimation result of SOC can greatly affect the other management of the battery.However,SOC is an internal state variable of the battery,and there is no way for the sensor to obtain the variable value directly.Therefore,this thesis explores various SOC estimation methods for single-liquid flow zinc-nickel battery,and finally chooses the unscented Kalman filter algorithm based on singular value decomposition(SUKF).By introducing the singular value decomposition(SVD)method into the covariance updating process of the UKF algorithm,this algorithm solves the problem that the state covariance matrix P of the traditional unscented Kalman filter algorithm(UKF)tends to be trapped in a non-positive definite and non-symmetric state,resulting in solution failure.Considering that the SUKF algorithm alone estimates the SOC ignoring the time-varying battery model parameters,this paper proposes to use SUKF in conjunction with other filtering algorithms to form a two-layer filtering estimation algorithm(e.g.,SUKF-EKF,SUKF-FRLS,DSUKF),which can jointly estimate SOC and battery model parameters.The effectiveness of the joint algorithm is tested by and variable current continuous condition,and the results indicate that the joint estimation algorithm is much lower than the SUKF estimation algorithm both in terms of mean absolute error(MAE)and root mean square error(RMSE).Since the estimation performance of the Kalman filter algorithm is largely affected by the initial values of the noise and state covariance matrix variables,however,the selection of the initial values of the matrix variables by the traditional method is selected empirically,which cannot maximize the estimation performance of DSUKF algorithm.Therefore,this thesis uses the Harris Hawk optimization algorithm(HHO),which has been highly applied recently,to optimize the initial covariance matrix parameters,and introduces a multi-strategy combination to propose an improved Harris Hawk optimization algorithm(IHHO),which strengthens the algorithm’s optimization-seeking capability.Meanwhile,this thesis introduces the capacity update formula into the SOC estimation algorithm and generates the improved two-layer SVD traceless Kalman filter algorithm(IDSUKF).The new estimation algorithm not only improves the SOC estimation accuracy,but also predicts the aging state(SOH)and realizes the joint estimation of SOC and SOH.更多还原