Risk Prediction of Cascading Failures in Renewable AC/DC Hybrid Power Systems Based on Data-Driven Methods

With the large-scale integration of renewable energy, the stability of AC/DC hybrid power grids is facing more severe challenges, and traditional risk assessment methods are no longer able to meet real-time decision-making needs. This article proposes an online risk prediction and dominant factor analysis method for cascading failures based on real-time operation data of the power grid using a Wide Area Measurement System (WAMS). It can accurately and quickly predict the failure risk of each equipment in each stage of the entire process of cascading failures in the power grid. Firstly, a cascading failure model for an AC/DC hybrid power grid with wind power was established. Secondly, using real-time measurements of system voltage, current, and power as inputs to the model, and based on the LightGBM method, constructing failure risk prediction models for each device in each stage. Finally, by combining the SHapley Additive exPlans (SHAP) method, the prediction results of the model can be analyzed to quantitatively evaluate the impact of different operating parameters on the risk of single equipment failure and identify their dominant factors. The proposed method balances prediction speed, accuracy, and interpretability, which helps to enhance the safe operation capability of new energy AC/DC hybrid power grids.