Power grid frequency prediction using spatiotemporal modeling. (6th July 2021)

• Record Type: Journal Article
• Authors: Lenzi, Amanda; Bessac, Julie; Anitescu, Mihai
• Other Names: Morris Max D. guestEditor.
• Is Part Of: Statistical analysis and data mining. Volume 14:Number 6(2021)
• Page Start: 662
• Page End: 675
• Publication Date: 2021-07-06
• Subjects: hierarchical models -- phasor measurement unit -- power systems -- spatiotemporal correlations; Data mining -- Statistical methods -- Periodicals; 006.312
• DOI: 10.1002/sam.11535
• Languages: English
• ISSNs: 1932-1864

Abstract

Abstract: Understanding power system dynamics is essential for interarea oscillation analysis and the detection of grid instabilities. The FNET/GridEye is a GPS‐synchronized wide‐area frequency measurement network that provides an accurate picture of the normal real‐time operational condition of the power system dynamics, giving rise to new and intricate spatiotemporal patterns of power loads. We propose to model FNET/GridEye grid frequency data from the U.S. Eastern Interconnection with a spatiotemporal statistical model. We predict the frequency data at locations without observations, a critical need during disruption events where measurement data are inaccessible. Spatial information is accounted for either as neighboring measurements in the form of covariates or with a spatiotemporal correlation model captured by a latent Gaussian field. The proposed method is useful in estimating power system dynamic response from limited phasor measurements and holds promise for predicting instability that may lead to undesirable effects such as cascading outages.