基于VMD-PSO-GRU模型的锂离子电池剩余寿命预测

Remaining life prediction of lithium-ion battery based on VMD-PSO-GRU model

准确预测锂电池的剩余使用寿命(remaining useful life,RUL)对降低电池使用风险和保证系统的安全运行起着非常重要的作用。为了消除电池容量序列受容量再生等影响,提高预测结果的准确性和稳定性,提出了一种基于变分模态分解(variational modal decomposition,VMD)与参数优化的门控循环神经网络(gate recurrent unit,GRU)相结合的RUL预测模型。首先采用VMD算法将锂电池的容量序列分解为一系列平稳分量;然后采用多层GRU网络对各分量进行预测,针对预测结果不稳定的问题,在模型训练前利用粒子群算法(particle swarm optimization,PSO)对GRU模型的参数进行优化;最后叠加各分量的预测值作为最终预测结果。在NASA数据集上对本模型进行了验证,当采用20个已知电池序列数据预测时,预测结果的最大平均绝对百分比误差和均方根误差控制在0.88%和0.0148以内,RUL预测的最大误差不超过2个充电周期,具有较高的鲁棒性和预测精度。

Accurately predicting the remaining useful life(RUL)of lithium batteries plays a significant role in reducing battery risk and ensuring the safe operation of the system.To reduce the influence of battery capacity regeneration,and improve the accuracy and stability of RUL prediction,an integrated prediction model of variational modal decomposition and gate recurrent unit(GRU)network with particle swarm optimization algorithm was proposed.First,the capacity sequence of the lithium battery was decomposed into a series of stationary components using the variational modal decomposition algorithm.Then,the multiple GRU network was used to predict the capacity component on each sub-sequence.To deal with inconsistent prediction results,particle swarm optimization was used to optimize the parameters of the GRU model before model training.Finally,the predicted results of each sub-sequences were integrated as the final battery capacity estimation,followed by the prediction of RUL.The experimental results on the National Aeronautics and Space Administration dataset showed that the maximum mean absolute percentage error and root mean square error of the results were controlled within 0.88%and 0.0148,respectively,and the maximum error of RUL prediction was less than two charging cycles.It has high robustness and prediction accuracy.