Improvement of electric vehicle safety using a new hybrid fuzzy Q-learning algorithm for lithium-ion battery state-of-charge estimation

In the context of electric vehicles, knowing the state of charge and the amount of remaining energy in the battery pack is crucial, as it contributes to supporting and establishing smart energy management systems in electric vehicles. However, it is usually challenging to obtain this information directly in such applications. To address this issue, a new innovative algorithm combining fuzzy logic and Q-learning has been developed to estimate the state of charge for a lithium-ion battery based on a second-order battery model. A more accurate and flexible state-of-charge estimation is produced as a result of this integration. The hybrid fuzzy Q-learning algorithm processes input variables including battery voltage, current, and temperature using a membership function-based fuzzy inference system. Fuzzy rules translate these inputs into language expressions and fuzzy sets, allowing the system to reason and decide based on partial and ambiguous data. The fuzzy inference system parameters are updated using Q-learning, which increases estimation accuracy by learning from prior experiences. This was done using MATLAB/Simulink software. The results obtained showed that the proposed strategy is capable of achieving the highest possible accuracy in estimating the battery state of charge. Therefore, this strategy can be effectively utilized in practical applications.