The task of document-level relation extraction aims to identify the semantic relationships between entities in a document. However, the common pronoun ambiguity problem in documents is often not fully considered, resulting in serious issues of unclear pronoun reference in the document itself. In addition, a group of entity pairs may contain multiple relationships, making the task of relation extraction more challenging. In this paper, we propose a new model, EPRRC, to address the above challenges by implementing pronoun resolution and enhancing relation correlation. Specifically, we construct prompts for Large Language Models (LLMs) to achieve pronoun resolution and enhance the completeness of entity information. Ultimately, we obtain the pronoun enhanced dataset PR-DocRED. Then, in the relation filter module, we effectively identify the potential multiple relations between entities based on prior knowledge of relation co-occurrences. The experimental results indicate that our proposed PR-DocRED dataset exhibits significant performance improvements across many model tests. Moreover, the EPRRC model outperforms numerous strong baselines in fundamental evaluations.