Frequency Characteristic Analysis and Extraction from Transmission Line Measurement Data

Introduction/Objective: Under the status of expanding scale and complexity of national power grids, frequency is distributed in spatiotemporality. This study analyzes and extracts these characteristics to enhance power grid modeling, security control, and operational analysis. Methods: Using extensive generation-transmission-demand measured datasets, we analyze the spatiotemporal features of frequency measurements. We use the frequency spatial correlation identification method, which is based on the Pearson correlation coefficient. Additionally, a frequency fingerprint extraction method using Convolutional Neural Networks (CNN) is developed to capture high-dimensional features. These features are from frequency domain representations. The method is tested on measured frequency signals from five cities, including Beijing and Changzhi. Results: The results reveal a clustering phenomenon in frequency data. This clustering phenomenon is related to Alternating Current (AC) power grid structures. The CNN-based fingerprint extraction method effectively identifies spatial correlations. As a result, it achieves high recognition accuracy (specific figures provided in the study). Discussion: The findings are similar to the existing research on power grid spatiotemporal dynamics. As a result, we introduce a novel CNN-based approach for feature extraction. Limitations include data variability across regions. In the future, we will do more to explore real-time applications and expand datasets for broader validation. Conclusion: This study successfully extracts and analyzes spatiotemporal frequency characteristics. Meanwhile, we have demonstrated the effectiveness of CNN-based fingerprinting and correlation methods. All these results contribute to improved power grid monitoring and control, and these will contribute to the smart grid optimization in the future.