Live update of the battery thermal parameters for reliable short-term temperature model

Battery Thermal Management systems are a key component for modern electric vehicles. Many systems use advanced models for temperature prediction and for the optimal cooling or heating strategies. Thermo-electric characterizations of cells and battery packs are performed inside the research labs to properly tune the models before uploading them to the on-board control units. However, these data depict the storage system under brand new conditions. Throughout the vehicle life, the battery pack behavior can change because of several factors related to the surrounding environment and operating conditions. Moreover, in the case of battery swap strategies, that consist in the pack replacement, the temperature model would be totally unsuitable for the new installed one. Therefore, an on-board, online procedure for the evaluation and update of the battery thermal behavior could be needed. This work presents a method for the evaluation of the battery thermal parameters on-board, during real driving cycles using data that are available in the vehicle control unit. The main novelty of this work consists in the solution to combine and assist the temperature model in the vehicle control units with an optimization algorithm which does not increase the computational load and provides reliable thermal parameters estimations. To demonstrate the potential of this methodology, the evaluated parameters are used in a short-term temperature model suitable for control strategies for battery thermal management systems. Concerning the first part, an optimization procedure is run for different driving cycles, recorded using a GPS system on a real vehicle. Finite-difference method is used to identify the convective heat transfer coefficient and the specific heat capacity of a single cell that composes the battery pack in laboratory tests. Then, the reliability of the estimated thermal parameters is analyzed reducing the number of source records and using the remaining cycles for the validation. Four worst cases have been identified and used to check the performance of the model prediction. Considering a temperature measure tolerance of 6 %, up to 93.75 % of the estimated values is reliable. Finally, the thermal parameters are used in a short-term temperature model for control strategies. The results highlight the good performance of the model in the estimation of the on-board battery temperature during a real driving cycle, simulating the future heat generation on the basis of the current load demand of the previous time step. The temperature predictions of the short-term model have been also tested in the worst cases denoting a good reliability; the maximum error is + 5 % in overestimation and 3 % in underestimation. Temperature predictions would help in the feed-forward control of battery thermal management systems for a smooth and safe operation of future electric vehicles. Moreover, this solution can be adopted for even more complex cooling methods, enhancing the update of the battery pack characteristics also in case of deterioration or battery swap.