In vehicular ad hoc networks (VANETs), content caching at edge devices can facilitate direct content delivery without fetching content from the remote server. The existing cache strategy regards all vehicles on the road as cache nodes with identical characteristics, without considering the differences among vehicle types. Buses on the road have larger cache space and more stable driving speeds and paths. By fully utilizing these bus features, cache performance can be enhanced. Therefore, we propose a hierarchical cooperative edge caching strategy based on Double Deep Q-Network (DDQN). In this scheme, buses with a large communication range, stable driving speed and paths are used, along with RSUs, form a high-level cache backbone network. This network provides content caching and delivery services for private cars at the lower level, which have a high privacy and a considerable degree of mobility randomness. At the same time, we design a cache replacement strategy based on Double Deep Q Network (DDQN) model to optimize the content access delay. Simulation results show that compared with the traditional benchmark scheme, the proposed scheme demonstrates a substantial enhancement in cache hit ratio and reduction in content access delay.