This research proposes a real-time situational portrait label generation method for command and control systems based on Large Language Models (LLMs), aiming to address the limitations of traditional state evaluation techniques in terms of real-time performance and intelligence. The method constructs a multi-level labeling framework composed of acquisition, statistical, and rule-based labels. It leverages the Qwen3 Large Language Model to achieve automatic extraction of resource and task state features and situational portrait construction. Furthermore, in the rule-based label generation process, CART decision trees and a LightRAG mechanism are innovatively introduced to improve the accuracy and interpretability of rule generation. Experimental results demonstrate that the method exhibits high-precision and robust label generation performance across various typical operational scenarios, validating its broad applicability and engineering value in complex distributed command and control systems.