At present, the State Grid Corporation has established a relatively perfect emergency plan system, but the emergency plan still stays on the “paper,” which is not conducive to the quick search for specific disposal measures, and the command personnel are more based on the past experience in the face of complex and changeable emergency scenarios. In order to realize real-time and instruction-oriented interaction between electric power emergency plan and emergency personnel, it is urgent to study the generation method of electric power emergency plan information interactive instruction. In the process of generating electric power emergency plan instruction, it is necessary to build the relationship between instruction template and labels according to the actual emergency types and related emergency plans. It is necessary to divide the emergency plan into sections, clauses and words, and extract the entity corresponding to the specific event type, stage, link, scenario, executor, trigger condition, prerequisite, recipient, task measure, resource demand, and other labels in the text. This paper uses computer natural language processing technology to realize the generation of electric power emergency plan instructions, improve the digital degree of emergency plan, and lay a foundation for the wisdom of emergency plan.