The remaining useful life prediction of supercapacitor is an important part of the supercapacitor management system. In order to improve the reliability of the entire supercapacitor bank, this paper proposes a life prediction method based on long short-term memory neural network. It is used to learn the long-term dependence of degraded capacity of supercapacitor. The Dropout algorithm is used to prevent overfitting and the neural network is optimized by the Adam algorithm. The supercapacitor data measured under different working conditions is divided into training set and predictive set as the input of the neural network. The root mean square error of the predicted result is about 0.0261. At the same time, in order to verify the applicability of the algorithm, it is also used for the life prediction of offline data, and the root mean square error is about 0.0338. The overall results show that long short-term memory neural network exhibits excellent performance for remaining useful life prediction of supercapacitor.

超级电容器的剩余使用寿命预测是超级电容器管理系统的重要组成部分。为了提高整个超级电容器组的可靠性，提出了一种基于长短期记忆神经网络的寿命预测方法。它用于了解超级电容器容量下降的长期依赖性。Dropout算法用于防止过度拟合，并且神经网络通过Adam算法进行了优化。在不同工作条件下测得的超级电容器数据分为训练集和预测集，作为神经网络的输入。预测结果的均方根误差约为0.0261。同时，为了验证算法的适用性，还用于离线数据的寿命预测，均方根误差约为0.0338。总的结果表明，长期的短期记忆神经网络在预测超级电容器的使用寿命方面表现出优异的性能。