Research on quick judgment of power system stability using grid hierarchy net

Abstract

Although deep learning has been introduced in the stability simulation analysis of power system, the structure of model needs to be further studied. A good structure can reflect the essence and simplify the solving process, like convolutional neural network (CNN) for image recognition. In this paper, a novel neural network model is proposed based on the power grid connection called grid hierarchy net (GHNet). The model can significantly reduce the number of trainable parameters while using more input variables to improve the accuracy of the model. Firstly, the construction method of GHNet is introduced based on electrical distance of stations. Then, some key issues are discussed including input and output selection. Finally, the actual data of the Northeast Power Grid of China was used to verify the feasibility and effectiveness of GHNet which meets the requirements of online security and stability analysis.

Keywords

• Power system
• Dynamic security assessment (DSA)
• Deep learning
• Model structure
• Grid hierarchy net (GHNet)