Deep learning architectures for inference of AC-OPF solutions
This work addresses the problem of accelerating optimal power flow computations for electrical grid operators, but it is incremental as it focuses on comparing existing architectures rather than introducing a new method.
The authors systematically compared neural network architectures for inferring AC-OPF solutions, demonstrating that leveraging network topology through graph-based models improves performance over fully connected baselines in regression and classification tasks, with computational gains for obtaining optimal solutions.
We present a systematic comparison between neural network (NN) architectures for inference of AC-OPF solutions. Using fully connected NNs as a baseline we demonstrate the efficacy of leveraging network topology in the models by constructing abstract representations of electrical grids in the graph domain, for both convolutional and graph NNs. The performance of the NN architectures is compared for regression (predicting optimal generator set-points) and classification (predicting the active set of constraints) settings. Computational gains for obtaining optimal solutions are also presented.