Abstract: With the increasing complexity of power grid structures and the rapid growth of power data volume, traditional data detection methods struggle to meet the real-time and accuracy requirements of power systems. An abnormal data detection algorithm based on a domestic supercomputing platform can provide high-performance computing support, improving detection efficiency and reducing errors in power grid data. This paper proposes a power grid abnormal data detection method based on a domestic supercomputing platform, combining deep learning models and a multi-layer integrated detection framework. A case study is conducted to address the efficiency and accuracy demands of abnormal data detection in power grids, demonstrating that parallel algorithm designs based on domestic supercomputing systems can effectively enhance detection efficiency.