Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Utilizing AMI Interval Data and Machine Learning Algorithms to IdentifyDistribution System Topology and DER Connectivity

Cook, Elizabeth (2021) Utilizing AMI Interval Data and Machine Learning Algorithms to IdentifyDistribution System Topology and DER Connectivity. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

Download (1MB) | Preview


The ongoing deployment of Distributed Energy Resources (DERs), while bringing benefits, introduces significant challenges to the electric utility industry, especially in the distribution grid. These challenges call for closer monitoring through state estimation, where real-time topology recovery is the basis for accurate modeling. With the dramatic increase of the residential photovoltaic (PV) systems (i.e., DER), utilities need to know the locations of these new assets to manage the unconventional two-way power flow for sustainable management of distribution grids. Previous methods to maintain the system connectivity are either based on outdated maps or an ideal assumption of an isolated sub-network for topology recovery, e.g., within one transformer. This requires field engineers to identify the association, which is costly and may contain errors. As it has been shown that, historical records are not always up-to-date.

To solve these problems, a density-based clustering method is proposed that leverage both voltage domain data from the Advanced Measurement Infrastructure (AMI) and the geographical space information. The goal of such a method is to efficiently segment data sets from a large utility customer pool, after which other topology reconstruction methods can carry over. Specifically, it is shown how to use the voltage data and GIS information to refine the connectivity within one transformer. To give a guarantee, a theoretic bound for the proposed clustering method is shown, providing the ability to explain the performance of the machine learning method. Numerical results on both IEEE test systems and utility networks show the outstanding performance of the new method. An implementation is also demonstrated in the field.
In this dissertation, we consider the rich potential of large utility datasets, in which physical laws are inherently embedded, to identify system information and utilization by using machine learning algorithms. In order to provide situational awareness and tackle practical issues such as limited measurements and un-scalability, we start with proposing a customized data-driven approach to provide an accurate model for distribution grid control and planning.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Cook, Elizabethelizabeth.cook13@gmail.comemc138
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee CoChairGrainger, Brandonbmg10@pitt.edubmg10
Committee CoChairWeng,
Committee MemberAkcakaya, Muratakcakaya@pitt.eduakcakaya
Committee MemberDalla, Ahmed HassanAHD12@pitt.eduAHD12
Committee MemberBarati, Masoudmasoud.barati@pitt.edumasoud.barati
Committee MemberKelly-Pitou,
Date: 13 June 2021
Date Type: Publication
Defense Date: 6 October 2020
Approval Date: 13 June 2021
Submission Date: 3 April 2021
Access Restriction: 2 year -- Restrict access to University of Pittsburgh for a period of 2 years.
Number of Pages: 101
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Electrical and Computer Engineering
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: distribution power grid, machine learning algorithms, system topology recovery, distributed energy resources
Date Deposited: 13 Jun 2021 18:44
Last Modified: 13 Jun 2023 05:15


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item