基于间接健康指标与回声状态网络的航空锂电池剩余使用寿命预测

Remaining Useful Life Prediction of Aviation Lithium Battery Based on Indirect Health Index and Echo State Network

航空锂电池是飞机的重要组成部分,对其进行剩余使用寿命(RUL)预测至关重要。目前,一般采用锂电池容量作为RUL指标,但在飞机实际运行中,锂电池容量难以准确测量,同时面临长期寿命预测精度的问题。因此,提出一种基于间接健康指标和回声状态网络(ESN)的锂电池RUL预测方法,基于一阶偏相关系数分析方法提取最能代表电池寿命的间接健康指标来代替容量指标,建立间接健康指标预测模型。同时基于深度学习中的ESN,结合粒子群优化算法(PSO)对网络参数进行优化,建立退化预测模型,实现锂电池RUL预测,解决长期预测精度问题。使用NASA电池数据进行实验验证说明,该方法相较于其他方法具有更高的精度、稳定性以及良好的泛化能力。

Aviation lithium battery is an important part of the aircraft.It is essential to research remaining useful life(RUL) prediction of lithium battery.At present, the capacity of lithium battery is generally used as the RUL index, but in the actual operation of the aircraft, the lithium battery capacity index is difficult to accurately measure, and it also has the problem of long-time life prediction accuracy.Therefore, a RUL prediction method based on indirect health index and echo state network(ESN)is proposed.Based on the first-order partial correlation coefficient analysis method to extract the indirect health index that can best represent the battery life to replace the capacity index, the indirect health index prediction model is established.At the same time, based on ESN in deep learning, combined with particle swarm optimization(PSO) to optimize the network parameters, a degradation prediction model is established to realize the RUL prediction of lithium battery, and solve the problem of long-term prediction accuracy.Experimental validation using NASA battery data shows that this method has higher accuracy, stability, and good generalization ability compared to other methods.