In order to accurately grasp the general law of the evolution of public opinion in extreme rainstorm disaster events, and to explore the evolution characteristics of public opinion of the government, the media, Opinion leader and Internet users in extreme rainstorm disaster events, a topic evolution model based on the co-word network and an emotion analysis model based on the emotion dictionary were constructed. Taking the April 11, 2019 rainstorm in Shenzhen as an example, the change process and differences of different subjects’ concerns, emotional attitudes, etc. were analyzed, Put forward management and control strategies for network public opinion in response to extreme rainstorm disaster events. Research has found that the four types of subjects have different focuses and significant differences in emotional attitudes; With the development of extreme rainstorm disaster, the attention content and emotional attitude of all subjects are constantly changing. Except the government, the attention of the other three types of subjects lags behind the development of the event itself. Negative emotions are mainly concentrated in the early warning period and emergency period, and then gradually tend to be neutral; The four types of subjects interact with each other. The government plays a leading role in the evolution of public opinion. The media is the follower of the government, and Opinion leader are more sensitive to the topics concerned by netizens. Finally, from the aspects of public opinion monitoring, public opinion guidance, information disclosure, response facilitation and so on, the network public opinion management and control strategy for extreme rainstorm disaster events is proposed.