# IN THE MAKING... CURRENTLY WORKING ON THE DATALOADER (DONE FOR TIMESERIES DATA)

IN THE MAKING... CURRENTLY WORKING ON THE MODEL!

Battery life cycle predition

Get the data

The data is available here. Download the three batches in .mat. Then create a data folder in the project's root directory and move the .mat there. The final folder structure should look like this:

long_cycle_life_prediction
├── data
|   ├── 2017-05-12_batchdata_updated_struct_errorcorrect.mat
|   ├── 2018-04-12_batchdata_updated_struct_errorcorrect.mat
|   └── 2017-06-30_batchdata_updated_struct_errorcorrect.mat

The data

The dataset is comprised of 124 batteries; each of them with a life cycle (a complete charge and discharge) range between 150 and 2300 approximately. It is split into 3 subdatasets:

• train (41 batteries)
• test (43 batteries)
• tets2 (40 batteries)

Batteries are charged and discharged until the battery reaches 80% of its original capacity. The number of cycles it takes for a battery to reach this state is called the battery use cycle. In the dataset this number lies between 148 and 2237 after filtering out the faulty batteries.

The data is comprised of scalar values associated to a specific cycle and the timeseries data corresponding to it.

The scalar values are comprised of:

• Internal resistance (Ω)
• Total charge (Ah)
• Discharge time (Minutes)
• Remaining cycles (Positive integer)

The timeseries data is comprised of:

• Temperature (°C)
• Charge (Ah)
• Voltage (V)
• Current (A)

Preprocessing

In order to preprocess the data run

python3 data_preprocessing.py

from the base directory. This will generate preprocessed_data.pkl file in your ./data directory.The generated result is a nested dict with the following structure:

dataset
    b1c1
        cycle_life # total cycles untill 80% of thenomial capacity has been reached
        summary # each of these subentries contain 1 value associated to each cycle 
            IR 
            QD
            Remaining_cycles
            Discharge_time
        cycles
            1
                Qdlin
                Tdlin
                Vdlin
            2
                Qdlin
                Tdlin
                Vdlin
                
            ...
            n
            
                Qdlin
                Tdlin
                Vdlin    
    b1c2
        ...
            ...