Emergency voltage control provides a real-time online response to maintain the long-term voltage stability of a power system. Searching for an optimal emergency control solution is a hard combinatorial optimization problem because of the highly dynamic and nonlinear characteristics of power systems with discrete control variables. Additionally, real-time response is required. A new mode-matching–based adaptive voltage control is proposed in this article to achieve a fast response by exploring local voltage profile modes and the technique of similarity ranking. First, some typical emergencies are studied in advance, and their control knowledge is collected for preparation. The set of local voltage profile modes and their corresponding optimal control solutions are stored as knowledge for each emergency. Second, fast online control is realized and the control solution is configured according to a similarity ranking among local voltage profile modes. Furthermore, a learning process is applied to accumulate knowledge in case of unprepared situations. Thus, improved control performances in the future are ensured. This article depends on a new concept of local voltage profile mode with which knowledge preparation, emergency identification, configuration of control, and knowledge accumulation work together to perform an effective response. Simulations of the IEEE 39-bus system verify the outcomes of this voltage control scheme.