Predicting AC Optimal Power Flows: Combining Deep Learning and Lagrangian Dual Methods

Authors: Ferdinando Fioretto, Terrence W.K. Mak, Pascal Van Hentenryck630-637

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The proposed model is evaluated on a large collection of realistic mediumsized power systems. The experimental results show that its predictions are highly accurate with average errors as low as 0.2%.
Researcher Affiliation Academia 1Georgia Institute of Technology, 2Syracuse University ffiorett@syr.edu, wmak@gatech.edu, pvh@isye.gatech.edu
Pseudocode Yes Algorithm 1: Learning Step
Open Source Code No The paper does not provide a direct link to a code repository or explicitly state that the source code for the described methodology is being released.
Open Datasets Yes The experiments examine the proposed models on a variety of power networks from the NESTA library (Coffrin, Gordon, and Scott 2014).
Dataset Splits Yes The experiments use a 80{20 train-test split and report results on the test set.
Hardware Specification No No specific hardware details (like GPU/CPU models, memory) are mentioned for running the experiments. The paper only mentions 'GPU memory' in the conclusion, but not as part of the experimental setup.
Software Dependencies No The paper mentions using the Adam optimizer, but no specific software names with version numbers (e.g., PyTorch, TensorFlow, or specific solvers like CPLEX) are provided.
Experiment Setup Yes The training uses the Adam optimizer with learning rate (α 0.001) and β values p0.9, 0.999q and was performed for 80 epochs using batch sizes b 64. Finally, the Lagrangian step size ρ is set to 0.01.