【摘要】 目的为模拟海洋工程与水下科考装备电池的真实使用情况,进行随机电流放电下的锂离子电池老化实验,通过高斯过程回归模型进行电池剩余寿命预测。方法从数据驱动方法中选取具备不确定性表达能力的高斯过程回归模型,选定核函数后通过训练数据来优化超参数建立预测模型。用随机应用下的电池充放电循环实验数据验证预测结果。结果与SE核函数相比,基于Matern核函数的模型预测效果更优。训练数据越多,预测起始点越大,模型预测绝对误差越小、MAPE与RMSE值更低。对两种不同实验温度、不同随机电流放电模式下的三组电池,模型预测绝对误差大多在40 cycle内,MAPE与RMSE值分别低于0.06、0.09,均能实现准确剩余寿命预测。结论对于随机应用下的锂离子电池剩余寿命预测,高斯过程回归模型具备高精度与适用性。更多还原

【Abstract】 Objective To simulate the actual use of battery in ocean engineering and underwater scientific research equipment, carry out the li-ion battery aging test with random discharging current and predict remaining useful life(RUL) through Gaussian process regression(GPR) model. Methods GPR model with uncertainty expression ability was proposed based on data-driven methods. After selecting the kernel function, the forecast model was established by training data to optimize hyper-parameters. The data set of charge/discharge tests of li-ion battery under randomized use was used to verify the prognosis results. Results Compared with SE kernel function, the GPR model with Matern kernel function could get better prediction results. The more the training data was and the larger the starting prediction point was, the smaller the absolute error was, the lower MAPE and the RMSE values were. As for three sets of batteries under two different temperature and two kinds of random discharging modes, the GPR model can get accurate prediction results. The absolute error was no more than 40 cycle and the MAPE and RMSE values were lower than 0.06 and 0.09 respectively. Conclusion The GPR model has high accuracy and strong applicability for RUL forecast of li-ion battery under randomized use.更多还原