東華大學圖書館 |

A Machine Learning-Based Approach for Synthetic Distribution Feeder Generation.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	A Machine Learning-Based Approach for Synthetic Distribution Feeder Generation./
作者:	Liang, Ming.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2020,
面頁冊數:	107 p.
附註:	Source: Dissertations Abstracts International, Volume: 82-09, Section: B.
Contained By:	Dissertations Abstracts International82-09B.
標題:	Artificial intelligence. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28225425
ISBN:	9798684618680

A Machine Learning-Based Approach for Synthetic Distribution Feeder Generation.
Liang, Ming.

A Machine Learning-Based Approach for Synthetic Distribution Feeder Generation. - Ann Arbor : ProQuest Dissertations & Theses, 2020 - 107 p.

Source: Dissertations Abstracts International, Volume: 82-09, Section: B.

Thesis (Ph.D.)--North Carolina State University, 2020.

This item must not be sold to any third party vendors.

Test systems are widely used by researchers and engineers to test conceptual designs, optimize parameter settings, and validate performance. However, developing high-fidelity distribution feeder models requires access to utility network models and customer data, which is a major barrier for the research community to have unrestrictive, unlimited number of customizable, realistic test systems for research and development purpose. So far, there has been very little attempt made towards the manual and static test system design principles, making creating an ensemble of test systems from actual feeder models a daunting task. Motivated by this, the dissertation aims at developing an end-to-end, machine-learning-based approach for automated, customizable test feeder generation using actual feeder models as inputs.This dissertation presents a novel, automated, generative adversarial networks (GAN) based synthetic feeder generation mechanism, abbreviated as FeederGAN. FeederGAN digests real feeder models represented by directed graphs via a deep learning framework powered by GAN and graph convolutional networks (GCN). Information of a distribution feeder circuit is extracted from its model input files so that the device connectivity is mapped onto the adjacency matrix and the device characteristics, such as circuit types (i.e., 3-phase, 2-phase, and 1-phase) and component attributes (e.g., length and current ratings), are mapped onto the attribute matrix. Then, Wasserstein distance is used to optimize the GAN and GCN is used to discriminate the generated graphs from the actual ones. A greedy method based on graph theory is developed to reconstruct the feeder using the generated adjacency and attribute matrices. After the feeder topologies and attributes are generated, we then use a statistical, rule-based method to generate the load transformers. The rules make the transformers follows line capacity constraints, line attributes constraints and topology constraints. Finally, we write the generated feeder file with load information into OpenDSS format and run combined case studies. The results show that the GAN generated feeders resemble the actual feeder in both topology and attributes verified by visual inspection and by empirical statistics obtained from actual distribution feeders.In the second part of the dissertation, a synthetic load generation method is developed via a novel sequential energy disaggregation (SED) algorithm. The SED algorithm is presented for extracting heating and cooling energy consumptions from residential and small commercial building loads using low-resolution (i.e. 15-minute, 30-minute, and 60-minute) smart meter data. The method is validated using data collected from 137 households in the PECAN street project. Results show that the proposed SED method is computationally efficient, simple to implement, and robust in performance. Based on the SED algorithm developed, case study is conducted on buildings with photovoltaic (PV) systems and electric vehicles (EVs). Among the cases, load database of zero-net energy (ZNE) cases, ZNE ready cases, ZNE with EV cases is built up. Therefore, those load data can be then used together with the generated synthetic feeders to test different planning and operational strategies.

ISBN: 9798684618680Subjects--Topical Terms:

516317
Artificial intelligence.
Subjects--Index Terms:

Machine learning

A Machine Learning-Based Approach for Synthetic Distribution Feeder Generation.
LDR:04485nmm a2200361 4500 001 2344608
005 20220531064606.5
008 241004s2020 ||||||||||||||||| ||eng d
020 $a 9798684618680
035 $a (MiAaPQ)AAI28225425
035 $a (MiAaPQ)NCState_Univ18402038269
035 $a AAI28225425
040 $a MiAaPQ $c MiAaPQ
100 1 $a Liang, Ming. $3 3683392
245 1 0 $a A Machine Learning-Based Approach for Synthetic Distribution Feeder Generation.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2020
300 $a 107 p.
500 $a Source: Dissertations Abstracts International, Volume: 82-09, Section: B.
500 $a Advisor: Chi, Min;Lubkeman, David;Baran, Mesut;Lu, Ning.
502 $a Thesis (Ph.D.)--North Carolina State University, 2020.
506 $a This item must not be sold to any third party vendors.
520 $a Test systems are widely used by researchers and engineers to test conceptual designs, optimize parameter settings, and validate performance. However, developing high-fidelity distribution feeder models requires access to utility network models and customer data, which is a major barrier for the research community to have unrestrictive, unlimited number of customizable, realistic test systems for research and development purpose. So far, there has been very little attempt made towards the manual and static test system design principles, making creating an ensemble of test systems from actual feeder models a daunting task. Motivated by this, the dissertation aims at developing an end-to-end, machine-learning-based approach for automated, customizable test feeder generation using actual feeder models as inputs.This dissertation presents a novel, automated, generative adversarial networks (GAN) based synthetic feeder generation mechanism, abbreviated as FeederGAN. FeederGAN digests real feeder models represented by directed graphs via a deep learning framework powered by GAN and graph convolutional networks (GCN). Information of a distribution feeder circuit is extracted from its model input files so that the device connectivity is mapped onto the adjacency matrix and the device characteristics, such as circuit types (i.e., 3-phase, 2-phase, and 1-phase) and component attributes (e.g., length and current ratings), are mapped onto the attribute matrix. Then, Wasserstein distance is used to optimize the GAN and GCN is used to discriminate the generated graphs from the actual ones. A greedy method based on graph theory is developed to reconstruct the feeder using the generated adjacency and attribute matrices. After the feeder topologies and attributes are generated, we then use a statistical, rule-based method to generate the load transformers. The rules make the transformers follows line capacity constraints, line attributes constraints and topology constraints. Finally, we write the generated feeder file with load information into OpenDSS format and run combined case studies. The results show that the GAN generated feeders resemble the actual feeder in both topology and attributes verified by visual inspection and by empirical statistics obtained from actual distribution feeders.In the second part of the dissertation, a synthetic load generation method is developed via a novel sequential energy disaggregation (SED) algorithm. The SED algorithm is presented for extracting heating and cooling energy consumptions from residential and small commercial building loads using low-resolution (i.e. 15-minute, 30-minute, and 60-minute) smart meter data. The method is validated using data collected from 137 households in the PECAN street project. Results show that the proposed SED method is computationally efficient, simple to implement, and robust in performance. Based on the SED algorithm developed, case study is conducted on buildings with photovoltaic (PV) systems and electric vehicles (EVs). Among the cases, load database of zero-net energy (ZNE) cases, ZNE ready cases, ZNE with EV cases is built up. Therefore, those load data can be then used together with the generated synthetic feeders to test different planning and operational strategies.
590 $a School code: 0155.
650 4 $a Artificial intelligence. $3 516317
650 4 $a Systems science. $3 3168411
650 4 $a Information technology. $3 532993
653 $a Machine learning
653 $a Synthetic distribution
653 $a Feed generation
690 $a 0800
690 $a 0489
690 $a 0790
710 2 $a North Carolina State University. $3 1018772
773 0 $t Dissertations Abstracts International $g 82-09B.
790 $a 0155
791 $a Ph.D.
792 $a 2020
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28225425