基于Markov链与SARIMAX模型的混合动力车辆速度预测
Speed Prediction of Hybrid Electric Vehicle Based on Markov Chain and SARIMAX Model

针对车辆速度预测较为复杂的问题，提出将马尔可夫模型与SARIMAX模型引入车辆速度预测方法之中。基于美国密歇根州安阿伯市的车辆数据，将汽车运行的速度数据处理为时间序列数据之后建立模型进行预测。详细对比分析了马尔可夫模型与SARIMAX模型在四种不同工况(低速、中速、中高速、高速)与不同预测长度(5个、10个、15个、20个)下预测车速的准确程度。经过研究可以得出：在低速工况下，利用SARIMAX模型和马尔可夫模型对车速进行短期预测或长期预测都是不错的选择；在高速工况下，长期预测应选择马尔可夫模型，短期预测两个模型都能获得较好的效果；在中速和中高速工况下，长期预测应选择SARIMAX模型，中速工况下的车速短期预测选择SARIMAX模型更合适，中高速工况下的车速短期预测，两种模型的预测效果欠佳。
Aiming at the complicated problem of vehicle speed prediction, Markov model and SARIMAX model are introduced to predict the vehicle velocity. Based on vehicle data of Ann Arbor, Michigan, USA, this paper processes the vehicle velocity data into time series data and then substitutes them into two models. The accuracy of Markov model and SARIMAX model in predicting vehicle speed under four different working conditions (low speed, medium speed, medium high speed and high speed) and different prediction lengths (5, 10, 15 and 20) is compared and analyzed in detail. Through practical research, it can be concluded that under high-speed conditions, Markov model is selected for long-term prediction, and SARIMAX model and Markov model for short-term prediction can achieve better results; SARIMAX model is selected for long-term prediction under medium speed and medium high speed conditions. SARIMAX model is more suitable for short-term prediction of vehicle speed under medium speed conditions. For short-term prediction of vehicle speed under medium and high speed conditions, the prediction effects of the two models need to be improved.