摘要
交通流的可预测性是进行短期交通流预测的基础。本文首先判别了短期交通流的混沌特性，求解出表征交通流“蝴蝶效应”的最大Lyapunov特征指数，在此基础上按照交通流动力系统运动轨道的演化特点求解出最大可预测时间，但是交通流系统是开放的复杂巨系统，最大可预测时间涉及到的影响因素很多，论文分析了交通流历史数据样本的大小和数据中含有的噪声对交通流可预测性的影响和随着预测步长的增加，交通流可预测性的衰减特征，得出交通流可预测性是一个综合指标，不能仅仅以最大Lyapunov指数的倒数来确定，应综合分析考虑。论文得到的结果在实际的交通流数据中得到了验证。
The predictability of traffic flow is the basis of short-term traffic flow forecasting. Firstly, the paper identified the chaotic character of short-term traffic flow. And then the largest Lyapunov exponent which reflects the ‘butterfly effect’ is solved. On this basis, the largest forecasting time scale could be found according to the evolvement of the traffic flow dynamic system. But the traffic flow system is an open, complex and huge system and there are a lot of factors that affect the predictability of traffic flow. The paper analyzes the effect of historic data scale and the noise in the original traffic data to the predictability of traffic flow. Also the paper studies the attenuation character of the predictability with the increase of the forecasting step. At last the paper concludes that the predictability of traffic flow is an integrated parameter. It could not be determined only by the reciprocal of largest lyapunov exponents. We should analyze it synthetically. The conclusion is validated in the real traffic flow data.
