Abstract: Technologies of large language models are not only a major breakthrough in the direction of natural language processing within the artificial intelligence field, but also transforming patterns of knowledge acquisition and innovation. The principles of knowledge acquisition and innovation of large language models are studied, their potential applications in the construction of information systems for emergency management are explored. It addresses the challenges currently faced in the intelligent emergency management and the limitations of intelligence level of the service systems. A knowledge management mode of reconfiguration intelligence emergency based on large language model technology is proposed. On such basis, the emergency management brain which can collaboratively innovate, percept in all-domain, and support decision-making is conceived so that the intelligence level of the whole service system can be improved from the perception intelligence to cognitive intelligence.