New Energy Battery Cabinet Discharge Resistance

Does a CC discharge voltage model lose quality at 1 min time step?

It is observed that the model yielded an MAE and RMSE of 65 and 83 cycles respectively at the minimum time step (4 s) whereas errors are 87 and 109 cycles respectively when the CC discharge voltage curves are sub-sampled at every 1 min. This indicates that the model does not lose significant quality at 1 min time step. 4.5.

Can machine learning predict battery capacity fade and internal resistance curves?

In this research, we propose a data-driven, feature-based machine learning model that predicts the entire capacity fade and internal resistance curves using only the voltage response from constant current discharge (fully ignoring the charge phase) over the first 50 cycles of battery use data.

How to construct features based on cc discharge voltage?

To construct features based on this we used statistical (mean, variance, kurtosis, skewness, min and max) and gradient based approaches (area under curve and slope). The calculation of gradient and area under the curve is not obvious and requires a careful processing of the CC discharge voltage and its corresponding time.

What is the relationship between IR and battery temperature?

This is most likely attached to the inverse relationship between this predictor and IR. As IR increases, the battery temperature rises and voltage drops thus a negative rate. The reverse is the case when the IR falls with time. Fig. 5. Pearson correlation between all generated features and targets. Fig. 6.