Research on electric vehicle state of health estimation based on multifeature attribute data mining

The inaccurate estimation of electric vehicle (EV) health status is a critical bottleneck restricting EV development. Accurate health status estimation is essential for consumers. This paper proposes a health state estimation method for EVs based on multi-feature attribute data mining. First, the method extracts battery-related health features to estimate battery health status. It employs four estimation algorithms and selects the gradient boosting regression model due to its lowest error, achieving RMSE and MAE values of 0.04496 and 0.03576, respectively. Additionally, it considers the impact of mechanical transmission efficiency, electric power transmission efficiency, and temperature on EV health status. Furthermore, it introduces a dung beetle optimization algorithm to determine weight coefficients and establish the final evaluation system, denoted as <i>y(x)=a*y1(x)+b*y2(x)+c*y3(x)+d*y4(x)+e*y5(x)+f*y6(x)</i>. The method applies the optimal coefficients obtained from the first eight vehicles to the ninth and tenth vehicles, resulting in RMSE values of 0.07683 and 0.07456, respectively.