Optimizing State Estimation Error with the LinDist3Flow Model
This paper was led by my undergraduate mentees Jeslyn Ero, Kieran Slattery, and Xianhe (Ken) Qin as part of the Georgia Tech Opportunity Research Scholars' Undergraduate Research Program.
BibTeX
@inproceedings{ero_lindist3flow_2025,
author={Ero, Jeslyn and Slattery, Kieran and Qin, Xianhe and Talkington, Samuel and Molzahn,
Daniel K.},
booktitle={2025 IEEE Opportunity Research Scholars Symposium (ORSS)},
title={{Optimizing State Estimation Error with the LinDist3Flow Model}},
year={2025},
volume={},
number={},
pages={1-4},
doi={10.1109/ORSS66051.2025.11121639}}Related Links
Abstract
This research paper presents an algorithmic approach to optimizing state estimation errors in unbalanced distribution networks as the integration of renewable energy sources such as solar and wind increases. These resources introduce uncertainty into grid models, challenging the satisfaction of engineering constraints.The study focuses on the impacts of this randomness on power flow equations and aims to enhance grid state estimation by combining electric grid physics with techniques from probability theory. The goal is to develop methodologies that effectively utilize uncertain measurement data, optimizing the accuracy and efficiency of smart meter data streams.The algorithm implementation aims to learn the optimal line parameters (resistance and reactance) in a three-phase unbalanced distribution system by minimizing the discrepancy between predicted and measured voltage values. It leverages the LinDist3Flow power flow approximation model to simulate voltage magnitudes, and formulates a loss function representing the squared error between the predicted and noisy measurements. The ADAM optimization algorithm iteratively minimizes this loss function, adjusting the parameters to achieve the best fit to observed data. This contributes to improving the accuracy of power flow modeling under noisy and unbalanced conditions.