The virtual simulation testing and verification of autonomous driving systems require large-scale test scenarios. Existing methods for scenario construction are often limited by human experience in user interaction and scene generation, leading to inefficiencies. To overcome this challenge, we propose a method that leverages LLaMA 3.1, an open-source large language model, within an intelligent collaborative framework to generate simulation scenarios using natural language commands. By integrating domain knowledge, R157 regulations, OpenX scenario samples, and other textual resources, the model generates text through training, fine-tuning, and quantization, with support for Chinese. It standardizes the generation of OpenX format simulation scenario files and references R157 regulation to automatically create cut-in scenarios. The application of these generated scenarios in simulation testing for the ALKS system demonstrates the effectiveness and feasibility of this scenario generation approach.