Remaining Useful Life estimation for electric vehicle batteries using a similarity-based approach

Following the rapid expansion of electrical vehicles, in the last few years many challenges related to lithium-ion batteries reliability and safety have gained substantial attention in the scientific community. In order to ensure a low life cycle cost and an optimized maintenance management it is fundamental to predict the future state of health of batteries installed in electric vehicles. In these regards, Remaining Useful Life estimation models allow to predict the battery’s end of life ahead of schedule with noteworthy accuracy results. In this paper, a similarity-based approach has been used to predict the future battery state of health based only on condition monitoring data acquired on similar devices. The proposed data-driven approach does not require a physic-based degradation model, allowing to easily extend the method to every kind of batteries available on the market. The effectiveness of the proposed method has been tested using the TRI (Toyota Research Institute) battery degradation dataset. A subset of the batteries has been used for training the model, while the remaining batteries have been used for validation. The results emphasize the accuracy of the proposed model and the ability to follow the non-linear battery degradation trend.