Developing a Predictive Model to Identify Potential Electric Grid Stress Events Due to Climate and Weather Factors

Monday, May 12, 2014 - 07:00
Add to Calendar

The electric grid is vulnerable to extreme climate and weather events because the impacts on electricity demand and/or supply resources can reduce reserve margins below target reliability thresholds. However, current metrics used to define events such as drought and heat waves have been developed largely for human health and agricultural purposes and may not be good predictors of electric grid stress events. To understand the potential increase in frequency and/or intensity of grid stress events under climate change, validated predictive models for grid stress events that address both demand- and supply-side impacts due to climate and weather are needed. This paper focuses on the first phase of this effort: the development and application of a robust predictive model specifically targeted at identifying the weather conditions leading to electricity demand stress. (Subsequent phases will address predictive models for supply-side stresses.) We have chosen the geographic region served by the Electric Reliability Council of Texas (ERCOT) region for our research. ERCOT tracks hourly electricity demand in each of 8 unique climate zones within Texas, providing a data-rich test-bed for model development. The analysis uses hourly demand data from 2003-2013 for each climate zone, as well as more limited availability hourly pricing data to categorize daily peak-hourly demand as stress and non-stress events. We then obtained coincident hourly weather station data (temperature, humidity, pressure, wind speed, and rainfall) from 50 unique weather stations within ERCOT to derive a representative hourly weather profile in each climate zone. From this hourly weather data, we derived multiple weather values (i.e. maximum, minimum, average) across varied time scales (i.e. daily, multi-day, and multi-hour) as potential covariates in each zone. We tested alternative predictive models across each climate zone during the historical period and chose the most robust for an analysis of future climate under RCP8.5, as simulated by a dynamically downscaled regional earth system model. We compare the frequency of future demand stress events identified by our new model as compared to the results of using a standard WMO heat wave definition.