Distribution Grid Topology Estimation Using 1D-CNN

Abstract

In order to improve the safety level and economy of distribution grid, a flexible and reconfigurable grid topology is the basic feature of future smart distribution grid. Many advanced functions of intelligent distribution grid, such as state estimation, power flow calculation and voltage control, require a correct grid topology as a prerequisite. Therefore, it is of great significance to find an accurate topology estimation method for distribution grid. This paper proposes a topology estimation method based on the one-dimensional convolutional neural network (1D-CNN). The proposed method firstly obtains the required voltage amplitude and phase angle data of distribution grid through simulation. The data should be standardized after the first step. Then the processed data is used to train the 1D-CNN model. Finally, the trained 1D-CNN model can be used for topology estimation. Compared to the existing methods, the proposed method does not need much historical data of each node in the distribution grid, and the calculation speed can support the online topology estimation. Meanwhile, this method is suitable for radial and weak loop grid structures. The effectiveness and superiority of the proposed method are validated by IEEE 33-node distribution grid.