*4.2. SOC Estimation Accuracy*

A rigorous analysis of SOC estimation accuracy performance can be performed using the information extracted from the SOC residual corresponding to the first scenario, Ro, i.e., for a SOCini value of 0.7 and all other parameters of Li-ion battery adjusted to the nominal values, as shown in Table 3. Moreover, the SOC accuracy is strongly related to the battery model accuracy. Since both Li-ion battery models are exactly accurate, as was shown in Part 1, an excellent e fficiency for all three estimators based on both battery models can be anticipated. The second assessment procedure of SOC estimation accuracy of each SOC estimator can also be carried out based on all six statistical criteria values obtained from Table 1. Moreover, a complete performance analysis consists of analysing the information provided by each SOC residual value and using statistical criteria. By inspecting the statistical criteria values, column by column, for each model, the AEKF SOC estimator based on the Simulink model behaves slightly better than two other competitors, followed by SOC estimators PFE and AUKF.


**Table 3.** The Li-ion SOC estimator accuracy based on the SOC residual error (%).

On the other hand, for a 3RC ECM battery model the AUKF behaves better, followed closely by AEKF and PFE. By far, combining the results obtained in Tables 1 and 3, it can be said that the AEKF

SOC estimator has better performance for the Simscape battery model, followed quite closely by the AUKF SOC estimator. For a 3RC ECM battery model it is the AUKF that performs better, followed by AEKF and PFE SOC estimators. However, since the values of statistical criteria extracted from Table 1 are close to each other for most of them, it is difficult now to make a net difference between the performance of all three SOC estimators. Moreover, sometimes it is difficult in some situations to make an interpretation that is approximative of each statistical criteria value. Still, in some cases, due to unsuitable values for the tuning parameters, the AEKF, AUKF and PFE SOC estimates are biased. Regarding all three SOC estimators, we observed that the SOC accuracy depends on a "trial and error" empirical adjustment procedure of tuning parameter values. Unfortunately, this procedure takes much time. Moreover, a new readjustment procedure is required when changing the driving conditions and SOC initial value, as well as when ageing and temperature effects take place. The adopted versions of AEKF and AUKF, due to their adaptive features, attenuate the tuning procedure of the parameters significantly.
