Battery Cycle Life Prediction

This project is for paper "A Hybrid Ensemble Deep Learning Approach for Early Prediction of Battery Remaining Useful Life"

Getting Started

The original data is available at https://data.matr.io/1/, download the data and put them into folder /Data

Prerequisites

1.After downloading the data, run BuildPkl_Batch1.py, BuildPkl_Batch2.py and BuildPkl_Batch3.py to extract the data for training and test

python BuildPkl_Batch1.py

2.Run Load_Data.py to delete bad battery data.

python Load_Data.py

Note: BuildPkl_Batch*.py and Load_Data.py are provided by author, small changes are made. Original code: https://github.com/rdbraatz/data-driven-prediction-of-battery-cycle-life-before-capacity-degradation

Feature Extraction

feature_selection.py provide the implementation of feature extraction and selection

Test with different models using feature_based model

feature_based_model_paper.py includes several model implementation with different features combination. --feature: variance, discharge, full --model: elastic,SVR,RFR,AdaBoost,XGBoost

python feature_based_model_paper.py --feature=0 --model=0

Proposed hybrid model with snaphsotensemble

pytorch_hybrid_model_snapshot_train.py includes hybrid model implementation. Before running, need to generate the dataset via data_process_for_hybrid_model.py.