短时交通流组合模型预测  

Short-term traffic flow forecasting based on hybrid model

作　　者：沈国江[1] 朱芸[2] 钱晓杰[2] 胡越[2] 

机构地区：[1]浙江工业大学计算机科学与技术学院,浙江杭州310023 [2]浙江大学工业控制国家重点实验室,浙江杭州310027

出　　处：《南京理工大学学报》2014年第2期246-251,共6页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(61174174);中央高校基本科研业务专项资金(2013XZZX008-1);浙江省交通运输厅科研计划项目(2014T08)

摘　　要：针对城市道路流量的非线性和不确定性特点,为避免单一模型预测准确率不高的缺陷,该文提出了一种短时交通流组合模型。该模型包含卡尔曼滤波模型和径向基函数神经网络模型2个子模型,较好地解决了神经网络不能反映大流量下的稳态性问题,以及卡尔曼滤波在流量不稳定时预测准确率不高的问题。在组合模型中引入惯性因子,确保了模型的稳定性。仿真结果表明该方法是可行有效的。In view of that the traffic flow of the urban road is a nonlinear and uncertaint interrupted flow,a hybrid model for the short-term traffic flow is put forward to overcome the shortage of the lower forecasting accuracy of the single model. This model consists of two sub-models,the Kalman filter model and the radical basis function neural network model,so the steady-state problem of the neural network model in the huge traffic flow and the low accuracy problem of the Kalman filter model in the unsteady traffic flow can be all solved. An inertia factor is introduced in the process of combining to ensure the stability of the hybrid model. The simulation result shows that the hybrid model is feasible and effective.

关 键 词：间断流 短时交通流预测 卡尔曼滤波模型 径向基函数神经网络 惯性因子 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]