Prediction of the remaining useful life of lithium-ion battery based on particle filter-long short term memory network

To the problem that it is difficult to accurately predict the remaining useful life (RUL) of lithium battery, a prediction model of improved long short term memory network based on particle filter (PF-LSTM) is proposed. The health factors extracted from the historical aging data of battery charge and discharge are selected as training samples which are closely related to the capacity decline issue of battery. The hyperparameters of LSTM including the number of neurons, learning rate, node abandonment rate, batch size, training steps, et al are optimized by PF algorithm. By the global optimization ability of PF, the prediction ability of the network is improved. Dropout layer is introduced to avoid network over-fitting, so the generalization ability of the model is improved. Experimental results show that PF-LSTM has the highest accuracy compared with other algorithms.