Searchable encryption plays a vital role in keeping data privacy of current cloud storage. Considering the security threat of trapdoor exposure, Emura et al. introduced a formal definition of keywords revocable public-key encryption with keywords search (KR-PEKS), which satisfies trapdoor exposure resistance (TER) and trapdoor re-generability. However, there is little work on lattice-based searchable encryption schemes to have the above properties, while it was widely accepted that cryptosystem on lattice has the potential to resist quantum computing attack. In this paper, we give a model of KR-PEKS with novel security, called bounded TER, i.e., a bounded number (previously set) of trapdoors for per keyword are exposed, which cannot affect other keywords. Then, we propose a lattice-based KR-PEKS scheme with bounded TER and trapdoor re-generability and show that this scheme is secure under the hardness assumption of learning with errors (LWE) problem. Finally, we conduct some simulation experiments on our scheme and give a comparison, which indicates that our scheme has less computational cost than the prior arts.