Table 2 Comparison of battery modeling, state estimation, and RUL prediction methods
From: Application of Digital Twin in Smart Battery Management Systems
Category | Methods | Advantages | Disadvantages |
|---|---|---|---|
Battery modeling | ECM | Simple parameter identification, small model computation and good real-time performance | Lack of physical meaning |
Electrochemical model | High precision, clear physical meaning | Large amount of calculation, model parameterization difficult, extreme conditions not applicable | |
DDM | High precision, suitable for dealing with nonlinear problems | Large amount of calculation, high dependence on training data | |
SOC estimation | Ampere-hour integral | Simple calculation, low cost, good real-time performance | High accuracy of sensors and accurate initial SOC |
Model-based method | High precision, strong adaptability and real-time performance | High dependence on models | |
Adaptive filtering method | Reducing the influence of sensor noise, high precision | High computational cost | |
Data-driven method | High precision, suitable for dealing with nonlinear problems | Large amount of computation | |
SOH estimation | SOC-SOH joint estimation | High precision | High dependence on models |
ICA/DVA | High precision, can react to the internal mechanism of the battery | The operation is difficult and time-consuming | |
Data driven method | The model is simple and suitable for different working conditions | A lot of data is needed, low efficiency of model updating | |
RUL prediction | Empirical prediction method | Simple process and less computation | Sensitive to the fluctuation of sample data, the results are easy to diverge. |
Filter prediction method | Reducing the influence of sensor noise, high precision | High dependent on the accuracy of empirical models | |
Time series forecasting | No need to consider the rationality of the model | High dependence of training data |