基于自适应神经模糊推理系统的锂电池荷电状态估计

Lithium Battery Charge State Estimation Based on Adaptive Neuro-Fuzzy Inference System

荷电状态(SOC)估计是保证锂电池可靠运行的一个关键环节。由于锂电池运行受到自身和外界诸多复杂因素的影响,传统的SOC估计方法效果不佳。因此,需要探索基于自适应神经模糊推理系统(ANFIS)的锂电池SOC估计方法。该方法利用电池电压、放电倍率、放电容量数据建立了三输入的ANFIS模型。ANFIS模型根据数据特征自行建立非线性模糊系统,并利用最小二乘估计(LSE)法实现模型结论参数优化,通过反向传播(BP)算法实现模型前提参数优化。将ANFIS模型与常用的BP模型进行比较,结果表明ANFIS模型相对于BP模型具有更好的估计准确率和稳定性,ANFIS模型的均方根误差(RMSE)值(7.09×10^(-5))和平均绝对误差(MAE)值(3.63×10^(-5))相对于BP模型,精度分别提高了1.1%和10.62%。

State of charge(SOC)estimation is a key process for ensuring the reliable operation of lithium battery.Since the operation of lithium battery is affected by many complex factors of itself and the outside,the traditional SOC estimation methods are ineffective.Therefore,there is a need to explore the SOC estimation method of lithium battery based on adaptive neuro-fuzzy inference system(ANFIS).The method establishes a three-input ANFIS model using measured voltage,discharge multiplier,and discharge capacity data.The ANFIS model establishes a nonlinear fuzzy system on its own according to the data characteristics,and optimizes the model's concluding parameters using the least squares estimation(LSE),and optimizes the model's premise parameters through the back-propagation(BP)algorithm.Comparing the ANFIS model with the commonly used BP model,the results show that the ANFIS model has better estimation accuracy and stability relative to the BP model,the root mean square error(7.09×10^(-5))and mean absolute error(3.63×10^(-5))of the ANFIS model improve the accuracy relative to the BP model by respectively 1.1%and 10.62%.