Prediction of Power System Frequency Stability Following Disturbance based on 2D-PCA and CNN

The integration of renewable energy sources and a high percentage of power electronic devices, combined with the hybrid connection of ultra-high voltage AC and DC, is the future direction of new power systems. However, the low inertia and weak anti-interference characteristics of these systems pose significant challenges to their frequency stability. Traditional methods for analyzing power system dynamics after disturbances are not sufficient in balancing speed and accuracy of calculations. By using the 2D-PCA method to reduce the dimensionality of large amounts of high-dimensional data, and then training a CNN offline on a training dataset and evaluating it online on a test set, the chosen algorithm has proven to provide quick calculation speeds and high precision. This enables operators to receive warnings of frequency instability risks during actual power grid operations, ensuring the system operates safely, stably, and efficiently. Additionally, after visualizing the data processed by the 2D-PCA-CNN model using t-SNE and comparing it with other deep learning algorithms, the effectiveness of the CNN based algorithm combined with 2D-PCA is clearly demonstrated.