Date of Award


Document Type


Degree Name


Organizational Unit

Daniel Felix Ritchie School of Engineering and Computer Science, Electrical and Computer Engineering

First Advisor

Amin Khodaei

Second Advisor

Aleksi Paaso

Third Advisor

Mohammad Matin

Fourth Advisor

Rui Fan

Fifth Advisor

Ryan Elmore


AMI, Distribution, Outage identification, Phase identification, Power system, Smart meter


In 2020, U.S. electric utilities installed more than 94 million advanced meters, which brought the percentage of residential customers equipped with smart meters to 75%. This significant investment allows collecting extensive customer data at the distribution level, however, the data are not currently leveraged effectively to help with system operations. This dissertation aims to use the smart meters’ data to improve the grid’s reliability, stability, and controllability by solving two of the most challenging problems at the distribution level, namely distribution network phase identification and outage identification.

Distribution networks have typically been the least observable and most dynamic and locally controlled elements in the power grid. Complete information about the network topology is continuously changing and is not always readily available when needed. Lack of phase connectivity information is a challenge, especially when rebalancing the grid and also in the aftermath of outages caused by extreme events. Traditionally, phase identification is executed manually. In this dissertation, a machine learning-based data mining method for accurate and efficient phase identification of residential customers is proposed by leveraging power consumption data collected through smart meters. The proposed method uses a high-pass filter to remove the redundant and irrelevant segments of the power consumption time series, and accordingly identifies the residential customers’ phase connectivity through a modified clustering algorithm.

Accurate connectivity information among customers is essential for outage identification and management in distribution networks. Extreme weather events can cause significant damage to electric power grid infrastructure and lead to widespread power outages. The frequency and the intensity of these events are continuously increasing as a direct result of climate change. Identifying grid components that are damaged is the first step to recovering from extreme weather-related power outages. An effective data mining method in identifying distribution network line outages is presented in this dissertation by leveraging data collected through AMI. The line outage identification method is developed based on a Multi-Label Support Vector Machine (ML-SVM) classification scheme that utilizes the status of customers’ smart meters as input data and identifies the outage/operational status of distribution lines.

Numerical simulations demonstrate the effectiveness of the proposed models and their respective viability in achieving the targeted operational objectives.

Publication Statement

Copyright is held by the author. User is responsible for all copyright compliance.

Rights Holder

Zohreh Sadat Hosseini


Received from ProQuest

File Format




File Size

80 pgs


Electrical engineering