fig2

Figure 2. The flowchart of battery health prognosis via EIS using the CNN-BiLSTM model. The inputs to our model include both the real (Z_re) and imaginary (Z_im) parts of impedance spectra collected at 60 different frequencies. The output is the capacity of a given cycle number or RUL. The dataset is partitioned into training data and testing data, and the parameters of the CNN-BiLSTM model are initialized. Subsequently, the training dataset is used to train the model. The predictive performance of the model is evaluated using two metrics: RMSE and R². Ultimately, well-trained models are obtained, enabling current SOH estimation, SOH forecasting and RUL prediction. EIS: Electrochemical impedance spectroscopy; CNN: convolutional neural network; BiLSTM: bidirectional long short-term memory; RUL: remaining useful life; RMSE: root mean square error; SOH: state of health.