Optimal Power Flow Calculation Based on SplitNN-DNN
MA Xing-he, Jie Wei, Zhihui Yi, Xizhe Li, Rui Song, Jian Chan · Journal of Physics Conference Series · 2024
Abstract Optimal power flow computation in power systems is an important aspect to ensure the smooth operation of power grids, but there are some challenges in terms of computing speed and data privacy protection in traditional methods. To address these issues, this study proposes a new method based on SplitNN-DNN. Firstly, we use a deep neural network to design a model that directly maps load data to voltage results, thereby improving computational efficiency. Second, we introduce a longitudinal federated learning technique to enable model training without data breaches. Finally, we conducted validation experiments on IEEE 118 and IEEE 300 node systems, and the results show that this approach not only protects data privacy, but also approaches the traditional centralized computation methods in performance, and has potential for practical applications.