Sensitivity Analysis of Battery Digital Twin Design Variables Using Genetic Programming

The advancement of digital twin (DT) technology improves battery performance and lifespan. Although precise forecasting, selection of design variables, and risk reduction are challenging. Therefore, it is critical in implementation of practical DT to investigate the sensitivity of feature implications on state estimation thoroughly. Hence in this paper, an analysis of features has been piloted using voltage and current characteristics. First, features have been extracted from performance values. Secondly, genetic programming (GP) has been set up to reflect the impact on state estimations. Structural risk minimization is used as a fitness function to maximize the DT's objective function, while GP-battery state estimation is implemented. An illustrative example is presented to evaluate the state of experimental data generated in the lab under controlled environmental conditions. Based on the analysis, the state of charge shows precision incorporation of all features, while the change in current over voltage shows the improvement in state of energy estimation. State of power is more sensitive towards changes in voltage concerning changes in current, and state of health offers better accuracy to the present voltage over the current applied. A sensitivity rating has been compared to design the role of the feature variable.