Battery State-of-Charge Estimation using RNNs

Code to train various models for state-of-charge estimation.

Data

This code is trained on the open source Panasonic 18650PF Li-ion Battery Data.

Code Setup

This code has the following dependencies:

• PyTorch
• Matplotlib
• Seaborn
• pandas
• mat4py
• scikit-learn

Our RNN script builds on code from https://github.com/yunjey/pytorch-tutorial/blob/master/tutorials/02-intermediate/recurrent_neural_network.

Preconfiguration

Create a data/ folder and extract the dataset mentioned in the Data section into it.

Train an MLP on the SOC Data

python mlp.py --window_size 10 --hidden_size 4 --noise_std 0.005

Train an LSTM on the SOC Data

python recurrent.py --model lstm --sequence_length 100 --hidden_size 4 --num_layers 4 --noise_std 0.005