Jolson Technologies is adept at a variety of battery modeling. Although our emphasis is on testing and data collection, models are a critical part of the science of battery performance. Models vary from extremely complex first principle models to empirical models based on data. We focus on empirical models based on neural networks. Neural networks were one of the first artificial intelligence algorithms and are particularly effective for determination of battery state of health and state of charge.
We have developed neural network models for both lithium-ion and lead-acid batteries with typical accuracy > 95%. We have shown that the data from a small cell can be used to model a larger cell with minimal loss in accuracy. We have focused mostly on state of charge (SOC) and state of health (SOH) but other variants are easily achieved.
The neural network models are ideal for battery management systems (BMS) because their output is one multiple input equation capable of fast computation. For a large pack, each cell must be evaluated independently, so a fast calculation is critical. In addition, they use data or derived metrics that can be obtained with the normal data acquisition of a BMS.
Please contact us if you would like to explore our modeling capabilities for your needs.