NREL: Wind Deployment System Model (WinDS): Background of Model

Qualitative details on wind intermittency

The intermittency of the wind resource can impact the electric grid in several ways. One useful way to examine these impacts is to categorize them in terms of time, ranging from multiyear planning issues to small instantaneous fluctuations in output.

At the longest time interval, a utility's capacity-expansion plans may call for the construction of more nameplate generation capacity. To meet this need, the planners can plan to build conventional dispatchable capacity or wind. The intermittency of wind precludes the planners from considering a MW of nameplate of wind capacity to be the same as a MW of nameplate of dispatchable capacity. The wind capacity cannot be counted on to be available when peak demand for electricity occurs. Actually, conventional capacity also cannot be considered 100% reliable. The difference is in the degree of reliability. Conventional generators experience forced outages on the order of 2%-20% of the time, while wind energy is available at varying levels that average about 30%-45% of the time depending on the quality of the wind site. For planning purposes, this lack of reliability is handled in the same way—a statistical treatment that calculates how much more load can be added to the system for each MW of additional nameplate wind or conventional capacity or Effective Load Carrying Capability (ELCC).

This Effective Load Carrying Capability is less for wind than for conventional capacity; first, because the wind availability is less than that of conventional generators. And second, because at any given instant, the generation from a new wind farm can be heavily correlated with the output from the existing wind farms—if the wind isn't blowing at one wind site, there is a reasonable chance it's not blowing at another nearby site. On the other hand, there is essentially no correlation between the outputs of any two conventional generation plants.

Fortunately, there are ways to partly mitigate both the low availability of the wind resource and its correlation between sites. In the past 20 years, we have seen considerable improvement in the wind capacity factors of new wind installations. This is attributable to both better site exploration/characterization and to improvements in the wind turbines themselves (largely higher towers).

The correlation in wind output between sites also can be reduced. Increasing the distance between sites and the terrain features that separate them reduces the chances that two sites will experience the same winds at the same time. Figure 5 shows this correlation as a function of distance between sites in both an east-west direction and a north-south direction. With its multiple regions, WinDS is able to approximate the distance between sites and, therefore, the correlation between their outputs. WinDS uses the correlation between sites to estimate the variation in wind output from the total set of wind farms supplying power to a particular region.

Figure 5. Correlation in Wind Output

Between each 2-year-period optimization and for each demand region, WinDS updates its estimate of the marginal ELCC associated with adding wind of each resource class in each wind supply region to meet demand within a NERC region. This marginal ELCC is a strong function of the wind capacity factor and the distance from the existing wind systems to the new wind site for which the ELCC is being calculated. It is also a weak function of the demand region's load-duration curve and the size and forced outage rates of the conventional capacity. This marginal ELCC is assumed to be the capacity value of each MW of that wind class added in the next period in that wind supply region to serve the NERC region demand.

Everything else being equal, when expanding wind capacity, WinDS will select the next site in a region that is as far from the existing sites as possible to ensure the lowest correlation and the highest ELCC for the next wind site. (More practically, everything else is never "equal," and WinDS considers the tradeoffs between ELCC and wind site quality, transmission availability/cost, and local siting costs.)

Generally, for the first wind site supplying a demand region, these capacity values (ELCC) are almost equal to the capacity factor. However, as the wind penetrates to higher levels, the ELCC can decline to almost zero in an individual wind supply region.

The next time frame of major interest is the day-ahead time frame. Utilities generally make decisions on which generating units to commit to generation a day ahead of time. To comply with these unit-commitment procedures, independent power plant owners may be expected to provide a bid for firm capacity a day ahead. Obviously, this can be problematic for intermittent-wind generator owners. For example, if the wind owner bids to provide firm capacity and the wind does not blow as forecast, the owner may have to make up the difference by purchasing power on the real-time market. If that power costs more per kilowatthour than he is being paid for his day-ahead bid, he will lose money on those kilowatthours he is forced to purchase.

Not all of today's electric grid systems operate day-ahead and real-time markets, as described in the preceding paragraph. For example, for wind, California allows a monthly balancing of bid and actual generation that is much more tolerant of the inaccuracies in forecasting wind a day ahead of time. However, in all cases, the imbalances can be offset with adequate operating reserves. Therefore, to capture the essence of the unit-commitment issue, WinDS estimates the impact of wind intermittency on the need for operating reserves (includes quick-start and spinning reserves) that can rapidly respond to changes in wind output. The operating reserves are assumed to be a linear function of the variance in the sum of generation (both wind and conventional) minus load. Because the intermittency of wind is statistically independent of the load variability and forced outages, the total variance with wind can be calculated as the sum of the variance associated with the normal (i.e., no wind) operating reserve and the total (over all the wind supply regions) variance in the wind output over the reconciliation period. Before each 2-year optimization, WinDS calculates the marginal operating reserve additions required by the next unit of wind (added in a particular wind supply region from a particular wind class) as the difference between the operating reserve required by the total system with that new wind and the operating reserve required by the total system, if there were no new wind installations in that wind supply region. This value is then used throughout the next 2-year linear program optimization as the marginal operating reserve requirement induced by the next MW of wind addition in that region of that wind resource class.

At the shortest time interval, instantaneous changes in wind output must be compensated for by regulation reserves. Regulation reserves are normally provided by automatic generation control of conventional generators whose output can be automatically adjusted to compensate for small changes in voltage on the grid. Fortunately, these instantaneous changes in wind output do not all occur at the same time, even from wind turbines within the same wind farm. This lack of correlation over time and the ease with which conventional generators can respond allows us to reasonably ignore this second order cost.

WinDS assumes that any wind generation delivered to a specific demand region in a specific time slice that exceeds the total electric load in that region/time slice will be lost. In addition, as mentioned above, WinDS also statistically accounts for surplus wind lost within a time slice due to variations in load and wind within the time slice.

WinDS has three endogenous options for mitigating the impact of intermittency. The first is to add conventional generators that can provide spinning reserve (e.g. gas combined-cycle) and quick-start capabilities (combustion turbines).The second, and usually least costly, is to allow the dispersion of new wind installations reducing the correlation of the outputs from the different wind sites. The third, and usually most costly, is to allow for storage of electricity at the wind site. If it is cost-effective, WinDS will build storage capacity at an onshore¹ wind site that can be used during peak electric demand periods to generate electricity when the wind is not blowing to its full capacity. The first storage concept examined in WinDS was the use of hydrogen. Thus, in the detailed equations and much of the discussion found in the next section of this report, hydrogen technologies (e.g., electrolyzers, fuel cells) are referred to interchangeably with storage concepts (e.g., electrolyzers equal conversion of electricity to storage, fuel cells equal conversion of stored energy to electricity).

WinDS endogenously selects the capacity of the turbines at a site, the transmission capacity to the site, and the capacity of the storage-conversion process (e.g., capacity of the electrolyzers required for hydrogen storage). It assumes onshore turbines are used to provide power either to the grid or to storage-conversion process (e.g., the electrolyzer). WinDS also assumes that power can be delivered to an onshore site from the grid (with industrial-power-purchase retail markups) to operate the storage-conversion process (e.g., electrolyzer) even when the onshore wind turbines are not generating power. This allows the capacity (and, therefore, the capital cost) of the storage-conversion process (e.g., electrolyzers) to be reduced, yet still provide the required energy (e.g., hydrogen) to storage (e.g., hydrogen). The energy stored at an onshore site is assumed to be used to power a generator (e.g., fuel cell) only during on-peak electric load periods.

WinDS also allows the electricity storage technology (e.g., electrolyzers) to be placed at the load centers rather than at the onshore wind site. General storage at the load centers would have greater capacity value than that located at an onshore wind site, because it would not be constrained by the availability of transmission capacity from the wind site. On the other hand, as mentioned above, storage at an onshore wind site allows for the downsizing of transmission-line capacity. In addition, there may be some capital cost synergisms with storage at a wind site—for example, a single control system might be used operate the wind and storage systems.

¹ To keep the number of variables in the LP to a minimum, storage at offshore wind sites is not simulated. Such systems are assumed to be less likely, due to the relatively higher costs of offshore wind and storage.