Post-Processing#

Most sources of battery data provide the voltage and current over time, but the other properties which are derived from them may be missing. The battery data toolkit provides “post-processing” classes which add compute these derived data sources.

All post-processing tools are based on the BaseFeatureComputer class and, as a result, provide a compute_features() function that adds new information to a battery dataset. Use them by first creating the tool and invoking that method with a BatteryDataset:

computer = FeatureComputer()
new_columns = computer.compute_features(data)

New columns will be added to a part of the dataset (e.g., the cycle-level statistics) and those new columns will be returned from the function.

The feature computers fall into two categories:

  • RawDataEnhancer, which add information to the raw data as a function of time

  • CycleSummarizer, which summarize the raw data and add new columns to the cycle_stats

Note

Post-processing is only supported for battdat.data.CellDataset for now.

Integral Quantities#

Functions which add columns associated with the accumulated values of data in other columns.

Time#

Compute columns which are derived fields associated with the relative time or timespans of data.