(37c) Rnn and Regression Analysis for Real-Road Driving Battery Life Prediction of Electric Vehicles Based on Obd Data and Periodic Inspection Measurements

Due to the rapid growth of the electric vehicle market, the global market for lithium-ion batteries is expected to increase by 27% annually (2018-2025), reaching $119 billion (about four times compared to 2018) by 2025. However, since the EV battery test and driving data are managed by the developer themselves and not disclosed, the existing battery remaining life prediction research is mainly based on open DB sources (NASA, Oxford, Stanford, etc.). In the case of Open DB, there is a limit to the use of accurate prediction of the remaining life of an electric vehicle as it is laboratory data that does not reflect actual road driving factors and battery aging factors. In addition, it is difficult to determine the residual value of electric vehicle batteries because there is currently no accurate method for predicting the remaining lifespan of used and rental electric vehicles.

In this study, a battery life prediction model was developed that solved the DB limitation of the existing electric vehicle battery remaining life prediction model. The developed OBD terminal can store data of more than 5000KM mileage through the SD card and can automatically collect real-time battery information based on real-time driving every 100ms. In addition, it is possible to store information of more than 100 vehicles such as electric vehicle cell voltage (96 cells or more), temperature, pack voltage, pack current, CCC, CDC, CEC, SoC, SoH, RPM, etc. For lifespan prediction based on the collected information, the remaining life expectancy model was developed using TensorFlow's LSTM and GRU. Basic factors such as pack current, pack voltage (96 cell voltages), module temperature, and the like, and a driving environment factor (instrument substrate capacity, battery pack external temperature, total charge/discharge value, etc.) and battery aging factor are used as input values, and accuracy of 90% residual life prediction was developed.