Development of an Online SOH Estimation Model Based on Mahalanobis Distance for Electric Vehicle Batteries

• Main Authors: LI, CHIA-YING, 李家瑩
• Other Authors: WANG,CHIEN-CHIH
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/gpy9nu

碩士 === 明志科技大學 === 工業工程與管理系碩士班 === 107 === In response to the global climate change and energy crisis, electric vehicles have gradually become the new craze of the market in recent years. State of health (SOH) is an indicator used to measure the excellent status of a battery. In the past, most studies on battery health were using laboratory equipment to measure parameters to estimate SOH before the battery left the factory, but there were few studies to estimate battery SOH after the sale. In order to know if the electric vehicle battery is healthy, the driver must return to the factory for battery testing. Because the battery test data is unbalanced data, this study proposes a method based on Mahalanobis distance to establish a model for the user to know the health of the battery after the battery is sold. This study validated the method with real data. The analysis results show that the original 15 commonly used measurement variables can be reduced to 2 key variables (rising voltage, discharge resistance). The predictive model has more than 90% classification accuracy to determine if the battery is healthy, indicating that the method can be effectively used for unbalanced battery data. In this study, SOH was estimated only by measuring the rising voltage and discharge resistance. These two variables are easily measured by the user, which is valuable for practical applications. We recommend that manufacturers can develop Bluetooth measurement equipment and APP in the future and provide an electric vehicle driving as a reference for replacing batteries or completing tasks.