基于流模型的缺失数据生成方法在剩余寿命预测中的应用

针对缺失数据生成模型精度低和训练速度慢的问题,本文基于流模型框架提出了一种改进非线性独立成分估计(Nonlinear independent components estimation, NICE)的缺失时间序列生成方法.该方法依靠流模型框架生成模型精度高、训练过程速度快的优势,并结合粒子群优化算法(Particle swarm optimization, PSO)优化NICE生成网络采样的退火参数,训练学习监测数据的真实分布,从而实现对数据缺失部分的最优填补.为进一步拓宽所提方法的应用范围,利用基于流模型的缺失数据生成方法得到的生成数据,通过建立融合注意力机制的双向长短时记忆网络(Bidirectional long shortterm memory with attention, Bi-LSTM-Att)的退化设备预测模型,实现设备剩余寿命的准确预测.最后,通过锂电池退化数据的实例研究,验证了该方法的有效性和潜在应用价值.

Missing Data Generation Method Based on Flow Model and Its Application in Remaining Life Prediction

Aiming at the problems of low accuracy and slow training speed of the missing data generation model, this paper proposes a missing time series generation method based on the flow model framework, which improves the nonlinear independent components estimation (NICE). The method relies on the flow model framework. The advantages of the high accuracy of the generative model and the fast training process are combined with the particle swarm optimization (PSO) algorithm to optimize the annealing parameters of the NICE generation network sampling, and the training and learning to monitor the real distribution of the data, so as to achieve the most optimal part of the missing data. In order to further broaden the application scope of the proposed method, this paper uses the generated data obtained by the missing data generation method based on the flow model, and establishes a bidirectional long short-term memory network (Bidirectional long short-term memory with attention, Bi-LSTM-Att) degradation device prediction model to achieve accurate prediction of the remaining life of the device. Finally, the effectiveness and potential application value of the proposed method are verified through a case study of lithium battery degradation data.