Frequency Stability Assessment Method Based on Wide-Area Spatial-Temporal Information Learning

Receiving-end power grids with heavy loads necessitate the delivery of large amounts of power via high-voltage or ultra-high-voltage AC/DC lines to meet local energy demands. In this circumstance, once DC blocking or line failure occurs, the receiving-end power grid would be easily affected by frequency dips. Moreover, the process of frequency drop is extremely evolved rapidly, often leading to irreversible frequency instability within several to tens of seconds. Therefore, it is crucial to rapidly and accurately predict the frequency instability trend of the receiving-end power grid. Considering this issue, this paper proposes a frequency stability assessment method based on wide-area spatial-temporal information learning. The method integrates a spatial interpolation algorithm and deep neural networks to build a transient frequency stability assessment model. The model assesses the system frequency stability status accurately, achieves high computational efficiency, and can meet the requirements for online use. A case study using data from an actual grid is conducted to verify the effectiveness of the method.