Real-time Power Transmission and Transformation Online Monitoring based on Convolutional Neural Network Algorithm

The convolution operation is vectorized by using single instruction multiple data stream (SIMD). The SIMD instruction in the CPU is used to realize the vectorized convolution operation, eliminating the synchronization overhead of parallel computing, and at the same time multiplexing the data in the register, reducing the number of memory reads and writes. Relying on the power production management information system (PMS), it aims to research and establish a set of A complete and unified online monitoring system for the status of power transmission and transformation equipment. The Internet of Things system architecture for power transmission and transformation equipment at the information integration layer and the intelligent application layer, and a detailed discussion of several key technologies involved in the system architecture, mainly including integrated intelligence Monitoring device, coding and identification system, communication technology, panoramic information modeling. The original input data is grouped and allocated to different data node groups to realize data parallelism; for the data of each data node group, pipeline parallelism is adopted.