hypochen / Battery-aging-modes-across-NMC

Battery aging modes across NMC

Battery-aging-modes-across-NMC

Introduction

Last updated by Bor-Rong (Hypo) Chen and Cody M. Walker.

Raw battery data and codes for battery aging mode classification.

The raw dataset consists of 44 NMC/Graphite single layer pouch cells. The data provided include cycle-by-cycle capacity, Coulombic efficiency, end of charge voltage (EOCV), and end of discharge voltage (EODV).

Summary of the cells

A summary of the 44 cells' information, including design parameters, cycling conditions, major aging modes, and experimentally obtained %LAM_PE, can be found in Pouch cell_summary.xlsx.

Battery data overview

Stored in the folder Battery raw data.zip.

Cycle-by-cycle battery data, including capacity, Coulombic efficiency, end of charge voltage (EOCV), and end of discharge voltage (EODV), are stored in folders named by the pack number and design:

• P462_NMC532_R2 design
• P492_NMC532_R1 design
• P531_NMC811_R1 design
• P533_NMC532_R2 design
• P540_NMC811_R2 design

(R1 = L_low and R2 = L_moderate design for electrodes)

The cycle-by-cycle data are in the format of .csv:

• capacity: Capacity_CellXX.csv
• Coulombic efficiency: CE_CellXX.csv
• End of charge voltage: (EOCV)EOC_CellXX.csv
• End of discharge voltage: (EODV)EOD_CellXX.csv

(XX indicates cell number)

Code overview

Download Battery raw data.zip and Pouch cell_summary.xlsx into a directory of your choice.

All of the codes used in data processing and analysis can be found in code folder.

Call Main_LLI_LAM_Classification.py and Main_LAM_estimation.py to process and analyze the battery data, including data grabbing and pre-processing, creation of a dataframe, data analysis, and plotting. Please change the file directory to fit your local file structure.

• Main_LLI_LAM_Classification.py will classify the cells into Li plating, SEI formation + less LAM_PE, and SEI formation + more LAM_PE.
• Main_LAM_estimation.py will perform a regression to estimate %LAM_PE.

The following is a library of codes that will be run by Main_LLI_LAM_Classification.py and Main_LAM_estimation.py.

• openPouchSummary.py selects the cells to serve as training data sets.

• fcnCBCdict.py grabs cycle-by-cycle data for each cell in each pack.

• detrendCBCdict.py removes spikes caused by RPTs in the raw battery data. This is done by treating them as seasonal effects and removing them with Seasonal Decomposition of Time Series with period.

• createDataframeFromPackDictV2.py finds trends within series to be used as predictor variables for Decision Tree Classification.

• createDataframeforLAM.pyis a replica of createDataframeFromPackDictV2.py, but includes the regression analysis of %LAM_PE.