No abstract

No abstract

Accurate estimation of the state of charge (SOC) is critical for the normal use of lithium-ion battery equipment like electric vehicles. However, the SOC of lithium-ion battery is not available by direct measure, but can only indirectly be estimated by measurable variables. According to the nonlinear characteristics between the measured values and SOC during the working period of lithium-ion batteries, we propose a method to estimate the SOC of lithium-ion batteries with Temporal Convolutional Network (TCN). The measured values of voltage, current, and temperature during the use of lithium-ion batteries can be directly mapped to accurate SOC in this method without using a battery model or adaptive filter. The network can self-learning and update parameters by being fed datasets collected under various working conditions and then obtain a model that can correctly estimate SOC under different estimation conditions. In addition, it can also be applied to different types of lithium-ion batteries through transfer learning with only a small amount of battery data. At various ambient temperature conditions, the average MAE estimated by the proposed method is 0.67% for all the tests, which proves that the TCN network is an effective tool to estimate the SOC of lithium-ion batteries.

In recent years, a notable development for predicting the remaining useful life (RUL) of components is prognostics that use data-driven approaches based on deep learning. In particular, long shortterm memory networks (LSTMNs) have been successfully applied in RUL prediction. However, to the best of our knowledge, these deep learning-based prognostics do not take into account uncertainty, and their prediction performance needs improvement. Bayesian model averaging (BMA) is a very useful ensemble method because it can quantify uncertainty. In this paper we propose a deep learning ensembled prediction approach based on BMA and LSTMNs. We constructed multiple LSTMN models with different subdatasets derived from the degradation of training data. Then, BMA was used to integrate the LSTMN submodels into one framework for a reliable prognostic. The main advantages of this method are that it 1) provides uncertainty management by postprocess forecast ensembles to create predictive probability density functions (PDFs) and generate probabilistic predictions with uncertainty intervals using BMA and 2) it improves prediction performance by ensemble multiple deep learning submodels (trained with different subdatasets) with corresponding weights calculated by the posterior model probability of the BMA. Finally, we introduced an online iterated training strategy for the BMA algorithm to realize higher prediction performance than that of an offline training strategy. In the experiments, we used lithium-ion battery data sets from the Center for Advanced Life Cycle Engineering at the University of Maryland. The results demonstrate the effectiveness and reliability of our proposed ensemble prognostic approach.