Research on power system frequency safety assessment method based on improved transformer

Abstract

With the ongoing development of power system and the widespread application of renewable energy, the complexity of power system in terms of electrical structure and dynamic characteristics is steadily increasing, presenting significant challenges to frequency safety. Accurately and efficiently assessing frequency safety is crucial, as it provides essential data support for emergency control strategies. Traditional assessment methods struggle to meet the demands for accuracy and responsiveness due to the intricateness of power system. However, advancements in artificial intelligence offer promising new avenues for frequency safety assessment. In this context, a model for assessing frequency safety in power systems is proposed, based on the improved Transformer. The proposed model enhances the Transformer’s feedforward network by integrating the convolutional residual structure of the Residual Neural Network (ResNet), thereby effectively combining the Transformer’s capability for sequential data processing with ResNet’s strength in feature extraction. This hybrid architecture significantly improves the overall assessment performance. Experimental evaluations on the IEEE 39-bus and Illinois 200-bus systems demonstrate that the proposed method surpasses conventional deep learning approaches in terms of accuracy, reliability, and false negative rate. Notably, the method provides rapid and precise frequency safety warnings, ensuring high accuracy and robustness.