Lithium Battery Life Prediction for Electric Vehicles Using Enhanced TCN and SVN Quantile Regression

The need for lithium battery life forecast algorithms has increased due to the global acceptance of electric vehicles. To further enhance the forecast’s precision and resilience about lithium batteries’ remaining life, this study implements quantile regression with in support vector networks to evaluate battery health conditions. Furthermore, it proceeds to integrate self-coding neural networks with temporal convolutional networks for the purpose of processing and extracting battery life data, and finally proposes a novel prediction model. The outcomes of the experiment demonstrate that when the width parameter is 0.75 and the penalty coefficient is 1, the battery health prediction accuracy of this new model is up to 88%, the remaining life prediction accuracy is up to 95.41%, and the number of battery capacity degradation times is up to 340, which is up to 45 times more than the number of the same type of model. In addition, the minimum difference in temperature prediction of battery charging under this model is close to 0.2°C, the minimum difference in temperature prediction during discharge is 0.3°C, and the battery capacity fidelity test findings’ average value is 91.08%. It is evident that the study’s suggested model offers a considerable advantage in estimating lithium battery lifespan for electric vehicles. Additionally, the study’s findings provide a quicker, more precise, and more flexible reference for estimating the lithium batteries’ condition and remaining life.