Riassunto analitico
The burgeoning interest in Artificial Intelligence (AI) and Machine Learning (ML) has spurred significant advancements in the exploration and development of novel techniques for assessing the state of batteries in electrified vehicles. Accurately estimating the State of Charge (SoC) of Li-ion battery cells allows for maximizing their utilization before disposal, thereby generating substantial cost savings in manufacturing and deployment. However, SoC determination remains a formidable challenge due to the limitations of current sensor technologies, which hinder accurate SoC measurements outside of laboratory settings. The main purpose of this work is to develop an automated pipeline for in-vehicle retraining of an existing battery AI model currently in use at the Hyundai Motor Europe Technical Center (HMETC). The development of this project is focused on defining the control logic that will govern the scheduler's behavior for in-vehicle training and the purge module that will handle the data and model databases according to appropriate logic. This process will be performed through advanced literature research and other solution comparisons to understand and implement the best possible control logic. The paramount objective of the automated pipeline is to alleviate the specter of range anxiety among drivers through the analysis of CAN-bus signals through a specific AI algorithm. The final segment of this work will include theoretical research for future pipeline development and improvement, with the aim of making the model as accurate as possible by introducing additional signals to the algorithm. In pursuit of this goal, we will investigate potential strategies for transitioning the pipeline to a cloud-based platform, thereby enabling the scalability of the battery AI model to serve the whole Kia and Hyundai fleet.
|
Abstract
The burgeoning interest in Artificial Intelligence (AI) and Machine Learning (ML) has spurred significant advancements in the exploration and development of novel techniques for assessing the state of batteries in electrified vehicles. Accurately estimating the State of Charge (SoC) of Li-ion battery cells allows for maximizing their utilization before disposal, thereby generating substantial cost savings in manufacturing and deployment. However, SoC determination remains a formidable challenge due to the limitations of current sensor technologies, which hinder accurate SoC measurements outside of laboratory settings.
The main purpose of this work is to develop an automated pipeline for in-vehicle retraining of an existing battery AI model currently in use at the Hyundai Motor Europe Technical Center (HMETC). The development of this project is focused on defining the control logic that will govern the scheduler's behavior for in-vehicle training and the purge module that will handle the data and model databases according to appropriate logic. This process will be performed through advanced literature research and other solution comparisons to understand and implement the best possible control logic. The paramount objective of the automated pipeline is to alleviate the specter of range anxiety among drivers through the analysis of CAN-bus signals through a specific AI algorithm.
The final segment of this work will include theoretical research for future pipeline development and improvement, with the aim of making the model as accurate as possible by introducing additional signals to the algorithm. In pursuit of this goal, we will investigate potential strategies for transitioning the pipeline to a cloud-based platform, thereby enabling the scalability of the battery AI model to serve the whole Kia and Hyundai fleet.
|