NEWS&EVENTS
News

Our faculty research achievements were awarded the best paper of IEEE Transactions on Power Systems


Recently, IEEE Power and Energy Society (Institute of Electrical and Electronics Engineers Power and Energy Society, IEEE PES ) has released its best papers for 2022. The paper "Data-Driven Optimal Power Flow: A Physics-Informed Machine Learning Approach " published by Postdoctoral researcher Lei Xingyu (first author), Professor Yang Zhifang (corresponding author) and Professor Yu Juan, etc., Department of Power Systems and Automation A Physics-Informed Machine Learning Approach, co-authored by Assistant Professor Junbo Zhao and others from the University of Connecticut, was awarded the best paper.

IEEE Transactions on Power Systems is the leading academic journal in the field of power system analysis and has important industry influence. The selection of the best papers includes all the papers published in IEEE Transactions on Power Systems in the past three years (a total of more than 1,400 papers), and a total of five papers were selected, with an acceptance rate of less than 4 percent.

 

 

Aiming at the problem that the existing data-driven optimal power flow calculation method has low applicability due to the difficulty of hyperparameter debugging and coordination, this paper proposes a physical information-guided data-driven optimal power flow calculation method by using a machine learning tool with simple structure and few tuning parameters. Firstly, a data-driven learning framework based on the physical characteristics of the optimal power flow is constructed, and the optimal power flow problem is divided into three stages of learning, which simplifies the learning difficulty of the optimal power flow problem and improves the calculation accuracy of the data-driven optimal power flow. On this basis, this paper proposes a pre-classification strategy for optimal power flow samples based on critical domain segmentation, which simplifies the complex mapping relationship between input and output of optimal power flow by pre-classifying samples with the same or similar constraints.

 

Since its official publication in 2021, the paper has been cited 37 times in Web of Science, 66 times in Google Scholar, and is a Popular paper in IEEE Transactions on Power Systems.

 

Link to the original paper

Data-Driven Optimal Power Flow: A Physics-Informed Machine Learning Approach

https://ieeexplore.ieee.org/abstract/document/9115822

 

Links to the original story:

https://cmte.ieee.org/tpwrs/tpwrs-best-papers/