Power System Overloading Risk Assessment Considering Topology and Renewable Uncertainties

This paper proposes a novel data-driven method for power systems overloading risk assessment considering topology changes and renewable energy uncertainties. By utilizing the Laplacian matrix and singular value decomposition, we reconstruct the original input by embedding the structural characteristics of the power grid within it. The topology embedding method presented in this paper offers a straightforward implementation process and avoids the need for parameter learning. The reconstructed input is further integrated with the partial parallel Gaussian process, which significantly enhances the computational efficiency of power flow simulations. The effectiveness of the proposed method is demonstrated on the modified IEEE 39-bus and actual ISO New England systems, via multiple criteria, i.e., mean absolute percentage error for point-wise accuracy, probability of exceeding the limit for quantifying risk, and expected excess for evaluating severity when an overload occurs.