Improving distribution system model accuracy by leveraging ubiquitous sensors

Power distribution;Power system modeling;Parameter estimation;Power system measurements;Power system simulation;Regression analysis;Smart grids

Peppanen, Jouni Aleksi ; Harley, Ronald G. Electrical and Computer Engineering Grijalva, Santiago Saeedifard, Maryam Meliopoulos, A. P. Ahmed, Shabbir ; Harley, Ronald G.

University:Georgia Institute of Technology

Department:Electrical and Computer Engineering

关键词: Power distribution;    Power system modeling;    Parameter estimation;    Power system measurements;    Power system simulation;    Regression analysis;    Smart grids;   

Others  :  https://smartech.gatech.edu/bitstream/1853/58179/1/PEPPANEN-DISSERTATION-2016.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

To enable advanced distribution automation schemes with ubiquitous distributed energy resources, it is imperative to increase distribution system modeling accuracy and detail, and to manage the Big Data from smart meters and other modern distribution system sensors. This dissertation shows accurate and computationally efficient parameter and topology estimation methods to calibrate existing utility distribution system secondary circuit low-voltage models with the modern distribution system sensor data. The methods are shown to be efficient with the Georgia Tech distribution system with smart meter measurements and with large utility feeder models with advanced PV inverter measurements. This dissertation also shows data validation and data imputation methods to manage the accuracy and reliability issues related to the modern distribution system sensor data. The presented data validation methods were effectively used to detect numerous issues in Georgia Tech smart meter data. Compared to conventional approaches, the presented data imputation method has a superior average accuracy in imputing Georgia Tech smart meter measurements. The method is computationally efficient and creates a series of imputed samples that have a continuous profile with respect to the adjacent available measurements, which is a highly desirable feature for time-series analyses.