Machine Learning-Driven Prediction of Load Shedding During Cascading Outages

The complexity of modern power systems makes predicting and mitigating cascading outages challenging. Quantifying their impact requires extensive simulations, which are computationally expensive and impractical for real-time applications. To address this issue, we introduce a novel methodology that integrates different machine learning regressors to create a stacking ensemble regressor along with a cascading failure model enabling rapid quantification of cascading-driven nodal load shedding through prediction. The effectiveness of the proposed model is evaluated using the R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> metric, with studies conducted on the IEEE 39-bus system. Additionally, the model is tested on the IEEE 24-bus Reliability Test System to demonstrate the effect of network reliability on cascading failures and its impact on model fitness. The outcomes of this study indicate that the proposed model fits better on the IEEE 39-bus system, achieving an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.88. In contrast, on the IEEE 24-bus Reliability Test System, the R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> is 0.42; the reasons behind this discrepancy are examined and discussed.