基于大数据和马尔科夫链的行驶工况构建

摘要 为提高代表行驶工况的准确性，对行驶工况构建算法进行了研究.在沈阳市选取10辆乘用车并采用自主驾驶方式收集行驶数据，组建了大样本数据库.首先根据傅里叶变换对原始数据进行了降噪滤波，然后采用改进的Kneser-Ney平滑方法计算状态转移概率矩阵，提出了基于马尔科夫链的行驶工况构建算法，最后开发了沈阳市乘用车代表行驶工况，并将其与数据库总体特征进行对比.结果表明，构建工况与数据库总体的平均偏差为2.46%，所有特征参数偏差均在10%以内，验证了算法的有效性.
Abstract：In order to improve the accuracy of typical driving cycle， the constructing algorithm for typical driving cycle was studied. 10 passenger cars in Shenyang City were selected to collect the driving data by autonomous driving， and the big sample database was established. Firstly， the raw data was filtered for noise reduction by using Fourier transform method. Secondly， the modified Kneser-Ney smoothing method was applied to compute the state transfer probability matrix， and the driving cycle constructing algorithm based on Markov chain was proposed. Finally， the typical driving cycle for passenger cars in Shenyang City was constructed and compared with the overall characteristics of the database. The results showed that the average deviation between the constructed cycle and the database population is 2.46%， the deviation values of all the characteristic parameters are within 10%， and the validness of the proposed algorithm is thus verified.