Retrieval-augmented generation for exam question creation in power industry education

As a fundamental sector of infrastructure, the power industry is characterized by its complexity and critical nature, which imposes high standards for talent development. Effective educational assessment tools are vital to maintaining the quality of training programs. However, traditional methods for constructing question banks rely heavily on manual input from experts, a process that is both time-consuming and labor-intensive, making it challenging to achieve comprehensive and diverse question coverage. To address these challenges, we introduce an intelligent question generation system that integrates Large Language Models (LLMs) with domain-specific knowledge bases, specifically designed for educational and training applications in the power industry. By employing Retrieval-Augmented Generation (RAG) techniques, the system extracts the most relevant information from the knowledge base, autonomously generates high-quality questions, and adaptively adjusts their difficulty levels. Experimental results indicate that the proposed method outperforms traditional approaches across multiple evaluation metrics, demonstrating its effectiveness and adaptability.