State of health (SOH) monitoring and remaining useful life (RUL) prediction are the key to ensuring the safe use of lithium-ion batteries. However, the commonly used models are inefficient in predicting accuracy and do not have the ability to capture local regeneration of battery cells. In this paper, a temporal convolutional network (TCN) based SOH monitoring model framework of lithium-ion batteries is proposed. Causal convolution and dilated convolution techniques are used in the model to improve the ability of the model to capture local capacity regeneration, thus improving the overall prediction accuracy of the model. Residual connection and dropout technologies are used to improve the training speed of the model and avoid overfitting in deep network. The empirical mode decomposition (EMD) technology is used to denoise the offline data in RUL prediction, so as to avoid RUL prediction errors caused by local regeneration. The proposed model is verified on two kinds of datasets and the results show that it has the ability to capture local regeneration phenomena in Lithium-ion batteries. Compared with the commonly used models, it has higher accuracy and stronger robustness in SOH monitoring and RUL prediction.INDEX TERMS Lithium-ion battery, state of health, remaining useful life, local capacity regeneration, temporal convolutional network.

No abstract

Accurate estimation of the state of health (SOH) of lithium-ion batteries is the key to ensure the safe use of lithium-ion batteries. In practice, the application of traditional health features is hindered by incomplete charge and discharge. When the battery is stably charged, the voltage and temperature of the battery under different health states show similar spatial degradation trends. Therefore, the degradation trend of voltage and temperature is directly taken as the health characteristic sequence through the dynamic time warping barycenter averaging (DBA) clustering. In addition, a new model attention depthwise temporal convolutional network (AD-TCN) considering health characteristics is proposed for SOH estimation. Depthwise separable convolution operation is used to extend temporal convolutional network (TCN) to a model suitable for multivariate prediction. Depthwise convolution is used as feature extractor, and pointwise convolution recombines all features for regression prediction. In addition, the convolutional block attention module is used in the channel dimension and spatial dimension to selectively enhance or suppress the details. Experiments on NASA data sets show that this method has strong reliability and high prediction accuracy.

No abstract

State of charge (SOC) is the most direct embodiment of the state of a lead-acid battery, and accurate estimation of SOC is helpful to ensure the safe use of the battery. However, the traditional estimation model has low precision and weak anti-interference. In this study, a new SOC estimation structure is proposed. This structure is based on the effective combination of the Isolation Forest (IF) anomaly detection algorithm and Long Short-Term Memory (LSTM) Network combined with Attention Mechanism (IF-LSTM-Attention). The Isolation Forest algorithm is used to effectively detect the missing values and outliers contained in the original data. Based on the actual charging and discharging data, a sliding window is constructed as the data of the model to give full play to the LSTM network length dependence. And LSTM network combined with Attention Mechanism achieves high-precision SOC estimation. In addition, the conventional dropout technique and Bayesian optimizer are used to improve the model training convergence rate. The results show that the IF-LSTM-Attention model proposed in this study has higher accuracy and better generalization ability than the conventional LSTM model and Back Propagation (BP) neural network model.INDEX TERMS Lead-acid battery, anomaly detection algorithm, long short-term memory network, state of charge.