【摘要】 为解决现有数据驱动的锂离子电池寿命预测方法中存在着精度低、训练数据需求大等问题，采用一种基于随机配置网络(SCN)的锂电池RUL预测方法。首先，以电池容量作为直接健康因子，电池放电电压作为间接健康因子，通过不等式约束分配随机参数，自适应选择随机参数的范围，建立锂电池剩余寿命的预测模型；其次，利用NASA电池数据集对预测模型进行训练；最后，将所建立的模型与FNN、CNN、LSTM等多种神经网络进行对比验证，充分发挥了SCN自主性强、收敛速度快、网络成本低等特点。结果表明，相较于其余神经网络，SCN的RMSE值最小，具有更低的训练损失和更好的网络拟合效果，是一种有效的锂电池RUL预测算法。更多还原

【Abstract】 In order to solve the problems of low accuracy and large demand for training data in the existing data-driven lithium-ion battery life prediction methods, a lithium-ion battery RUL prediction method based on random configuration network(SCN) is adopted in this paper.Firstly, taking the battery capacity as the direct health factor and the battery discharge voltage as the indirect health factor, the prediction model of the remaining life of lithium battery is established by assigning random parameters through inequality constraints and adaptively selecting the range of random parameters; Secondly, the prediction model is trained by using NASA battery data set; Finally, the established model is compared with FNN,CNN,LSTM and other neural networks, which gives full play to the characteristics of strong autonomy, fast convergence speed and low network cost of SCN.The results show that compared with other neural networks, SCN has the smallest RMSE value, lower training loss and better network fitting effect.It is an effective RUL prediction algorithm for lithium battery.更多还原