基于分布式最小二乘法的锂离子电池建模及参数辨识

精确的锂离子电池模型对于电池状态的准确估计以及电动汽车整车的仿真、设计与优化具有至关重要的意义。然而，传统的递推最小二乘方法应用于电池这类多时间尺度系统时，会出现模型参数辨识精度低、建模效果差等问题。为此，以锂离子电池二阶RC等效电路模型为研究对象，提出一种基于分布式最小二乘的模型辨识参数方法。此方法根据电池不同时间尺度可以分离的特性，将电池模型细分为两个子模型分别进行辨识，避免了待估参数的相互干扰，因而能够取得更好的参数估计效果。试验结果表明，相比传统的递推最小二乘辨识方法，提出的方法在UDDS和FUDS工况下能够将平均绝对误差分别降低约50.00%和28.57%，均方根误差分别减小约46.43%和29.17%，验证了所提方法的有效性和可行性。

Lithium-ion Battery Modeling and Parameter Identification Based on Decentralized Least Squares Method

An accurate lithium-ion battery model is of critical importance for the estimation of the battery states, and the simulation, design and optimization of electric vehicles(EVs). Low accuracy of model parameters and poor modeling performance occur when the conventional recursive least squares method(RLS) is used to estimate model parameters of lithium-ion batteries which show the multi-time scale characteristics. Therefore, a decentralized least squares method(DLS) is proposed and used to identify model parameters of the second-order RC equivalent circuit. The underlying principle of the proposed method is estimating model parameters of two sub-models separately by using the information that different time scales of batteries can be separated, and the estimated parameters can be obtained accurately because of the elimination of the mutual interference. The experimental results under UDDS and FUDS tests show that compared to RLS, the mean absolute error and the root mean square error of the proposed method can be reduced by about 50.0% and 28.57%, 46.43% and 29.17%, respectively. The effectiveness and feasibility of the proposed method are validated.