System and method of estimating state of charge of battery using coulomb counting estimation with periodic calibration using artificial intelligence based estimation

A method of tracking a state of charge of a battery which comprises a plurality of cells, comprising:

• counting coulombs for determining a charge difference over time for tracking changes of a state of charge of each of the plurality of cells of the battery and updating a corresponding plurality of state of charge values;
• and
• periodically invoking an artificial intelligence based estimator for converting a set of sample values of a selected one of the plurality of cells to calibrate the selected cell and to update a corresponding one of the plurality of state of charge values, wherein said periodically invoking an artificial intelligence based estimator comprises:
• converting scaled input vector data values into a plurality of state vector elements using a bidirectional Long Short-Term Memory layer;
• forming a linear combination of the plurality of state vector elements;
• and
• multiplying each of the plurality of state vector elements by a corresponding one of a plurality of weight values and combining resulting factors into a final state of charge value used to update a corresponding one of the plurality of state of charge values.

A method of tracking a state of charge of a battery which comprises a plurality of cells, comprising:

• counting coulombs for determining a charge difference over time for tracking changes of a state of charge of each of the plurality of cells of the battery and updating a corresponding plurality of state of charge values;
• and
• periodically invoking an artificial intelligence based estimator for converting a set of sample values of a selected one of the plurality of cells to calibrate the selected cell and to update a corresponding one of the plurality of state of charge values, wherein said periodically invoking an artificial intelligence based estimator comprises:
• converting scaled input vector data values into a plurality of state vector elements using a single Long Short-Term Memory layer;
• converting the set of sample values into a voltage variance value and a voltage mean value;
• forming a linear combination of the plurality of state vector elements, the voltage variance value and the voltage mean value;
• and
• multiplying each of the plurality of state vector elements, the voltage variance value and the voltage mean value by a corresponding one of a plurality of weight values and combining resulting factors into a final state of charge value used to update a corresponding one of the plurality of state of charge values.