Due to traffic congestion and environmental pollution, connected automated vehicle (CAV) technologies based on vehicle-to-vehicle (V2V) and vehicle-to-infrastructure communication (V2I) have gained increasing attention from both academia and industry. Connected hybrid electric vehicles (CHEVs) offer great opportunities to reduce vehicular operating costs and emissions. However, in complex traffic scenarios, high-quality real-time energy management of CHEVs remains a technical challenge. To address the challenge, this paper proposes a hierarchical eco-driving strategy that consists of speed planning and energy management layers. At the upper layer, by leveraging the real-time traffic data provided by vehicle-to-everything (V2X) communication, dynamic traffic constraints are predicted by the traffic flow predictor developed based on the Hankel dynamic mode decomposition algorithm (H-DMD). Then, the vehicle speed curve is planned under dynamic traffic constraints through a model predictive control (MPC) framework that takes fuel consumption, ride comfort, and traffic efficiency into account. To reduce the computational burden, an approximate vehicle fuel consumption model is introduced, where the fuel consumption is approximated by a function of vehicle power demand. At the lower layer, control-oriented models of fuel consumption and emissions are introduced. By tracking the optimized vehicle speed trajectory, the optimal power distribution among multiple power sources is calculated by the Pontryain’s Minimum Principle (PMP) under the MPC framework. Simulations have shown promising results. Under the constraints of a dynamic traffic environment, the proposed hierarchical eco-driving control strategy reduces fuel consumption and emissions per unit distance by 8.98% and 4.17%, respectively. Moreover, the planned speed profile with a small jerk (-0.1-0.1 m/s3) can substantially improve the ride comfort.