Power grid outage analysis improved using single deep neural network
AC Power Flow Contingency Analysis Using a Single Deep Neural Network
Machine Learning
Summary
Checking the safety of power grids when lines fail is important but slow using traditional computer models. The authors present a way to use one deep learning model trained only on normal conditions to quickly predict how the grid behaves after any single line fails. They show a mathematical way to ensure the model’s predictions will settle to the right answer, and tests show it works well on a standard test power grid. This method can speed up assessing grid reliability under different outage scenarios without retraining for each failure.
What this means in practice
- •For power system operators: Quickly estimate grid operating states after single-line outages without retraining models for each contingency.
- •For grid simulation tool developers: Incorporate a single neural network model to speed up contingency analysis across many outage scenarios.
Authors
Md Obaidur Rahman, Junjie Qin, Vassilis Kekatos
Abstract
Contingency analysis using the AC power flow (AC-PF) model is a critical tool for accurate grid security assessment, but its computational burden increases with the number of operating scenarios and outage configurations to evaluate. Recent ML-based approaches typically require outage-specific training data, leading to offline training costs that scale with the number of contingencies. This work proposes a framework that reuses a single ML model trained solely on basecase AC-PF data to estimate post-contingency operating states under arbitrary single-line outages. The proposed approach formulates post-contingency state prediction as a fixed-point iteration. If the ML model is a deep neural network (DNN), we derive sufficient conditions that guarantee convergence and develop semidefinite programming (SDP) formulations to certify these conditions for a given DNN. Numerical tests on the IEEE 118-bus system demonstrate that the proposed SDP formulations are tight, that the certified conditions hold for all tested contingencies, and that the resulting method produces accurate post-contingency state estimates within only a few iterations.