Presentation Details
| A Data-Driven Method for Synthetic Extreme Weather Generation and Solar Impact Assessment Duc-Huy Pham1, 2, Cong Feng1, Jin Tan1. 1National Laboratory of the Rockies, Golden, CO, USA.2North Carolina State University, Raleigh, NC, USA |
Abstract
High-resolution weather data are essential for assessing power-system resilience under extreme events, yet most existing extreme-weather datasets are localized, sparse, and insufficient for large-scale grid studies. This work proposes a data-driven synthetic extreme weather generation framework that enables flexible creation of realistic hazard scenarios for power system impact analysis. The method constructs a Weather Impact Factor (WIF) matrix by taking the ratio between gridded extreme-event weather fields and corresponding normal-condition fields derived from a reference dataset. The WIF preserves multivariate physical relationships among meteorological and solar variables and can be spatially manipulated to reposition the event over a desired study region without distorting feature dependencies. To support high-resolution applications, the transformed WIF is downscaled/interpolated to match a higher-resolution baseline dataset, and a nearest-neighbor mapping ensures full spatial coverage. The final synthetic extreme dataset is obtained by multiplying the adjusted WIF with the baseline normal weather, optionally scaled by a user-defined intensity factor. The framework is demonstrated using a hurricane scenario based on Hurricane Kathleen, reconstructed from ERA5 reanalysis and mapped onto the WECC 240-bus test system, while the NSRDB provides the 2-km, 5-minute solar baseline needed for detailed PV modeling. Solar generation impacts are quantified using the SAM–reV modeling pipeline, which models PV capacity factors and generation. Results show substantial disruption to solar output when the hurricane is directed toward Southern California, with system-wide solar generation drops by 50% during the peak interval, while California-region solar decreases by 69%, driven by large reductions in GHI due to storm cloud cover. Across WECC subregions, simulated solar output falls to 36–49% of clear-sky levels, consistent with published hurricane-related PV impact ranges. Overall, the proposed approach provides a scalable tool for generating diverse, synthetic-yet-realistic extreme weather scenarios to support solar-rich power system stress testing and data generation for machine learning applications.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
Content Locked. Log into a registered attendee account to access this presentation.