基于电池能耗的新能源汽车里程预测

New Energy Vehicle Mileage Forecast Based on Battery Energy Consumption

摘要: 行驶里程作为新能源汽车的一项重要评价指标，一直备受关注，对其里程进行高精度预测是当今电池能量管理系统领域的研究热点。本文基于北方某地区车辆1~4月实际行车数据，划分充放电片段，对各项参数进行数据清洗、相关性分析，充分考虑各项电池能耗相关参数特征，估算出车辆的平均能耗，进而得出车辆的最大估计里程和最小估计里程，同时，结合总电压数据，我们采用KNN算法对行驶里程进行了回归预测。结果发现，考虑电池各项参数对新能源汽车里程预测具有较为理想的效果，通过模型训练测试验证后其均方误差(MSE)稳定在3.6左右，R²达到0.97。具有较高的预测精度。

Abstract: As an important evaluation index of new energy vehicles, mileage has always been a concern. A study hotspot in the field of battery energy management systems is the high-precision prediction of mileage. Based on the actual driving data of vehicles in a certain area in the north from January to April, this paper divides the charging and discharging segments, performs data cleaning and correlation analysis on various parameters, fully considers the characteristics of various battery energy consumption related parameters, estimates the average energy consumption of vehicles, and then obtains the maximum estimated mileage and minimum estimated mileage of vehicles. At the same time, combined with the total voltage data, we use the KNN algorithm to predict the mileage. The results show that considering the parameters of the battery has an ideal effect on the mileage prediction of new energy vehicles. After model training and test verification, the mean square error (MSE) is stable at about 3.6, and R² reaches 0.97. It has high prediction accuracy.