Power Flow for Grids with Unfixed Topologies by Using Deep Neural Network

The power flow problem is critical in power systems, providing fundamental and vital information in system planning, operational management, and control strategy. The mathematical model of power flow problem is composed of several non-linear equations conventionally solved by the Newton-Raphson method. In this paper, the deep neural network which is called stacked auto-encoder (SAE), has been used to solve the power flow problem. Compared to the traditional neural network, stacked auto-encoder is a more effective method in terms of getting better local optimum and solving the problem of gradient diffusion. The resulted deep neural network can deal with the grids with various topologies corresponding to various line or bus faults. After a detailed description, the proposed method is evaluated by a comparison with a traditional neural network by Matlab simulation.