Lesson planning is essential to curriculum development but often imposes a heavy workload on teachers. This paper introduces an automated lesson plan generation framework based on a large language model (LLM), aiming to enhance efficiency and consistency. The framework comprises two core modules: a knowledge point extension module and a structured prompt engineering module, which collaboratively improve output relevance and structure. A custom dataset of lesson plans across six middle school subjects was constructed to support model training. Experimental results on five evaluation metrics demonstrate the framework’s superiority in generating high-quality lesson plans efficiently.