This paper aims to explore the fine-tuning application paradigms of Large Language Models and Large Multimodal Models in the field of industrial electrical automation. Firstly, it analyzes the issues of data complexity and knowledge barriers faced by industrial electrical automation systems in operation and maintenance, diagnosis, design, and scheduling. Secondly, it systematically discusses the technical paths of pre-trained model fine-tuning based on Domain Adaptation, including full-parameter fine-tuning, Parameter-Efficient Fine-Tuning (e.g., LoRA), and Prompt Engineering. On this basis, the paper constructs fine-tuning application models for core scenarios such as intelligent equipment fault diagnosis, production energy efficiency optimization, safety procedure generation, and assisted development of control codes. Furthermore, it elaborates in detail on key technical challenges and solutions, such as industrial data preprocessing, domain knowledge injection, and lightweight model deployment.