Convolutional Neural Networks : Encoding Battery Information

  • Tech Stack: Pytorch, Pandas, Matplotlib, Scipy

Known classification mechanisms include convolutional networks, but since we are dealing with time series, it is recommended to find image representations for each of the features.

A signal transformation technique, yields images that can be convoluted to classify the labels of each state of charge.

Classification of the State-of-Health becomes necessary in a different domain than just time, in order to obtain a feature map that can be trained to identify in which cases the batteries will reach their second life state of use. This makes the convolutional classifier a powerful tool in identifying defective batteries.