A Semantic Segmentation Algorithm for Distributed Energy Data Storage Optimization based on Neural Networks

There are many kinds of energy data, how to realize unified storage, processing and sharing of energy data is a big problem. As the national energy data center, State Grid aims to build a database that can store distributed heterogeneous asynchronous energy data. The storage of image files in the big energy database will take up a lot of space in the system, but not all parts of the image are needed. Therefore, it is very necessary to accurately segment the effective area of the image to store it so as to achieve the purpose of data compression. This paper proposes the Attention U-Net framework, which combines the traditional semantic segmentation network U-Net with the Attention module to focus on the region of interest in the image, emphasize foreground information, and suppress background information. The results show that compared with U-Net, the accuracy is improved by 1.77% and after the segmentation is completed, each image saves an average of 2MB of storage space.