Year: 2025
Type: article
Abstract: With the growth of global energy demand and the increasing complexity of the power market, the stable operation of the power grid is of utmost importance. This paper focuses on the artificial intelligence-based methods for power grid stability analysis. It details the basic principles, structural characteristics, and the error backpropagation algorithm of the fully-connected neural network, and explores the dynamic adjustment strategy of the learning rate to train a model with better performance. By testing the performance of models under different data partition widths, different numbers of neural network layers, and different activation functions, the results show that using the Sigmoid function as the activation function and medium- or wide - type neural networks contribute to improving the model performance. High-precision and high - recall models are trained to complete tasks with high requirements such as power system stability prediction.
Author: Shengze Xu
Institution: Guangxi University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4409777908