Modeling Variable Renewables in FACETS

Regions in FACETS, based on  nREL REEDS model  regions

Regions in FACETS, based on nREL REEDS model regions

An important issue in the modeling of deep decarbonization scenarios is the treatment of renewable resources like wind and solar. Generation from these resources varies over much finer time and geographical scales than multisector optimization models can represent. Capacity expansion models like FACETS generally break up geography into regions and the year into time slices. There are tradeoffs for model size in terms of numbers of regions and time slices. The current version of FACETS has 134 power regions, 356 wind resource regions, and 24 time slices representing 4 seasons and 6 times of day. 

Each small block shows the degree of correlation (red) or negative correlation (Blue) between load, PV, and wind curves in regions within ERCOT. AGGREGATION to model time slices worsens the apparent mismatch between load and wind generation and obscures spatial diversity in renewable generation, suggestion that the aggregation is a conservative approach to estimating renewable generation value.

Each small block shows the degree of correlation (red) or negative correlation (Blue) between load, PV, and wind curves in regions within ERCOT. AGGREGATION to model time slices worsens the apparent mismatch between load and wind generation and obscures spatial diversity in renewable generation, suggestion that the aggregation is a conservative approach to estimating renewable generation value.

A well-crafted time slice approach can capture the consistent expected patterns of hourly and seasonal variation in wind and solar generation. FACETS time slices were selected based on analysis of hourly historical load and generation patterns in wind and solar resource regions across the country to capture the times of mismatch between generation and load profiles. For example, in many regions, wind generation greatly exceeds load in the spring and fall pre-dawn hours. And of course, solar generation peaks during the summer mid-day hours, while loads peak several hours later. These mismatches can lead to curtailment if the excess generation cannot be economically stored or exported, comprising the value of adding further wind or solar to the system.

A correlation analysis was used to check that aggregation of curves into model time slices does not overstate the availability of generation to meet load. In fact, it shows that this aggregation worsens the mismatch between wind generation and load and obscures the diversity in wind generation timing at different locations, so that the model will understate the value of renewable generation, rather than overstate it.

What time slices cannot capture is the weather variations that lead to daily deviations in wind and solar generation from their seasonal time slice averages. Dispatchable resources, which can include storage, flexible load, and inter-regional trade, as well as dispatchable generation, are necessary to make up for these fluctuations. From historical hourly generation profiles, we estimated the expected over/under-generation across days in each of our seasons. Across regions in Texas and the Upper Midwest, the maximum seasonal expected deviation was around 25 percent. Based on this analysis, we've imposed constraints requiring a minimum 25 percent share of load in each time slice in each reserve region to be met by a combination of dispatchable generation, storage, and imports.

Variability also comprises the capacity value of wind and PV installations, that is, the degree to which their capacity can be counted upon in calculating the total regional capacity needed to meet peak load and maintain reserve capacity against unforeseen outages. FACETS draws estimates of the decline in wind and PV capacity values with increasing penetration from a multi-model study on modeling variable renewables by NREL, US DOE and EPA, and the Electric Power Research Institute. Capacity values decline to 5 percent for PV and 15 percent for wind, meaning that the model also must build backup capacity when investing heavily in wind and PV resources.

In evaluating the potential of renewable generation to meet load in future decarbonized systems, it is important to keep in mind that there are fine-grained tradeoffs between spatial and temporal variability that no optimization model can currently fully analyze. Operational experience over the coming decades will help to clarify any practical or economic limitations to renewable penetration. For now, our approach in FACETS has been to model the economics of spatial and temporal tradeoffs as best we can, while being conservative in representing possible limits to very high renewable penetrations.