Development of an intelligent method based on machine learning for prognostic and state of health estimation of electric vehicle batteries

In this doctoral thesis, the general ideologies of prognostic and diagnostic systems were examined. The machine learning models used in prognostic prediction methods and the challenges encountered in data-driven prediction methods were investigated. The study focused on lithium-ion batteries, exploring battery aging mechanisms and aging modeling. Data-driven modeling techniques and health status prediction studies developed using open-source datasets in the literature were reviewed. As part of the thesis work, eight lithium-ion cells used in electric vehicles were aged in a laboratory environment. Using the datasets obtained from this process, predictions for remaining useful life (RUL) and prognostic health status were made using Support Vector Machine (SVM) and Long Short-Term Memory (LSTM) methods. In the final section of the thesis, the predictive performance of SVM and LSTM methods was compared, and it was discussed that the LSTM method is more successful than the SVM method for prognostic health status prediction. The primary objective of this thesis is to develop an intelligent system based on machine learning that accurately predicts the health and lifespan of electric vehicle batteries, ensuring their safe and efficient operation throughout their service life. The studies conducted aim to enhance the performance and reliability of electric vehicles by monitoring battery health and detecting potential failures in advance.