For the purpose of accurately capturing the traffic state on expressways, this paper develops a multistep prediction algorithm of traffic flows on consecutive sections of expressways based on the state space model. First, the Van Aerde fundamental flow diagram is selected to replace the trapezoid structure utilized in the original cell transmission model (CTM), and its applicability under unequal cell lengths is discussed. Second, based on the modified cell transmission model (VA-CTM), the state space model is constructed by taking density as the state variable, and space mean speed and flow as output variables. In this process, we modify the state equation in consideration of the correlation between state and output noise vectors. Subsequently, a multistep prediction algorithm is designed based on the extended Kalman filter. Finally, several consecutive sections on the Western 3rd Ring-Road Expressway of Beijing are selected as the case study site to validate the proposed model. The data used in the case study include floating car data (FCD), remote traffic microwave sensor (RTMS) data, and the surveyed data. Predicted traffic states by the proposed model are compared with those by the Historical Average (HA) model and auto-regressive moving average (ARIMA) model. Results show that the proposed model outperforms HA and ARIMA models, especially in predictions of traffic states at bottleneck segments. This means that the proposed model is especially robust in depicting queues’ forming, dissipating, and shockwave.

CICTP