1.竞赛题目 Question

电动汽车行驶SOC预测

Prediction of SOC in the driving process of electric vehicles 

2.背景介绍 Background

由于电动汽车动力电池技术的限制，目前电动汽车的使用存在续驶里程不足，充电时间长，剩余续驶里程估计不准确等不便，这些缺陷在一定程度上阻碍了电动汽车的推广应用。电动汽车能耗的准确预测不仅能够帮助优化电动汽车控制策略从而减少能耗延长续驶里程，同时也能够为电动汽车剩余行驶里程的准确预测提供支持，进而为用户提供准确实用的驾驶辅助信息，帮助缓解里程焦虑。

Due to the limitation of battery technology, electric vehicles have limited endurance mileage, long charging process and difficulty in remaining driving mileage prediction, which have become the main obstacles in the application of EVs. The accurate prediction of electric vehicle energy consumption can not only help optimize the electric vehicle control strategy to reduce the energy consumption and extend driving mileage, but also, it can provide support for the accurate prediction of the remaining driving mileage of electric vehicles, and provide drivers with accurate and practical driving assistance information to help alleviate the mileage anxiety.

3.问题描述 Description of the question

动力电池能耗受累计行驶里程、外界温度和道路情况等多因素耦合影响。参赛者要设计动力电池能耗预测模型，对行驶过程的动力电池能耗进行预测。在本题中，车辆在一段行驶中的能耗水平通过SOC变化表征，参赛者需要根据行驶片段的初始SOC以及行驶过程中的相关参数预测行驶片段内SOC的变化情况。

注：SOC以车辆车载终端上传数据为准。

Power battery energy consumption is affected by the multiple factors such as cumulative driving mileage, ambient temperature and road conditions. Contestants are required to design a power battery energy consumption prediction model to predict the energy consumption of the battery during the driving process under real-world driving condition. In this question, the energy consumption during driving process is indicated by the change of SOC, contestants are required to use the initial SOC and related parameters to predict the SOC change.

Note: The SOC is subject to the uploaded data of the vehicle terminal.

4.数据说明 Data description

本赛题数据来源于市面上某款电动汽车的实际运行数据，训练样本和测试样本数据由同一车型的不同车辆产生。

训练样本包含5辆电动汽车的实际运行数据（含行驶和充电等状态），数据按时间顺序排列成行，列名见表1。数据可能存在异常，需参赛者自行识别。

The real-world driving data of electric vehicles are used in this question, and the training data and test data are produced by different vehicles with the same model. 

The training data consist of data of 5 vehicles (both driving state and charging state are included), and the data are arranged in time order in rows. The data items are listed in Table 1. Abnormal data may exist in the raw data and need to be identified by the contestants themselves.

测试样本由15辆车的150个行驶片段构成，每辆车10个行驶片段，测试样本中的0-4号车与训练样本中的0-4号车一一对应，5-14号车没有对应的训练样本数据。测试样本数据按时间顺序排列成行，列名见表2。数据可能存在异常，需参赛者自行识别。

Test data consist of 150 driving fragments of 15 vehicles, each vehicle provides 10 driving fragments. The vehicles No.0-4 in test data are corresponding to the vehicles No.0-4 in the training data, and vehicles No.5-14 have no corresponding training data.  The test data are arranged in time order in rows, and the data items are listed in Table 2. Abnormal data may exist in the raw data and need to be identified by the contestants themselves.