Transient voltage disturbance pattern recognition faces two challenges. First, disturbance recognition limited to a single monitoring point cannot accurately explain the complete process of complex disturbance, and the other is that off-line analysis cannot meet the requirements of real-time decision-making. Based on the Storm streaming computing framework, combined with logstash and Kafka message middleware, this paper builds a real-time data monitoring and processing platform for multiple monitoring points. The sliding time window algorithm is used to implement the Storm programming logic topology. By setting the basic time window size and quantity, it realizes the identification of transient voltage disturbance patterns for multi-temporal scale and multi-service models for regional power grids. The experimental results show that the number of tasks for properly setting the Storm component can maximize the parallel processing capability. The throughput and average processing delay results obtained by the simulation data test can meet the requirements of high throughput, scalability, real-time and accuracy of real-time processing of grid convection data. This work is supported by Fundamental Research Funds for the Central Universities (No. 2018MS030) and Science and Technology Project of State Grid Corporation of China “Big Data Deep Analysis and Application Technique Research for Power Quality of Urban Power Grid” (No. 52094018001C).