Capacity Loss Modeling of Li-Ion Battery Using Lightweight Neural Network Considering Equivalent Circuit Model

The research about battery health and degradation model has been extensively studied by researchers. The most accurate one that was recently developed by many researchers is using the data-driven method. Despite its high accuracy, the data-driven method implemented in degradation modeling prediction requires huge computing resources. So, when it is implemented on an embedded system, it will require a high specification of the controller and high cost. To overcome the problems above, a Degradation modeling based on equivalent circuit model and lightweight neural network is proposed. The battery is modeled on the equivalent circuit using the first order Thevenin model. From this equivalent circuit model, several parameters including internal resistance, open circuit voltage, and R-C voltage were obtained using a lightweight neural network model which uses 300 cycles of data of lithium ferrous phosphate (LFP) battery acquired from the experiment. These parameters will be obtained by fitting the battery testing data with minimum root-mean-squared error (RMSE) between terminal voltage from dataset and model output voltage. From this model capacity loss in every cycle can be predicted by fitting internal resistance obtained from neural network training with degradation data acquired from the experiment through the linear least square method. The result shows that the capacity degradation model using the proposed method can obtain 93.45% accuracy compared to actual degradation with small computational resources and minimum parameters from lightweight neural network model that consist of only two neurons on one hidden layer. So, this method can provide a lightweight battery degradation model with acceptable accuracy.