Grid Impacts of Wind Power Variability: Recent Assessments from a Variety of Utilities in the United States

This paper was presented at the 2006 European Wind Energy Conference (EWEC)

Selected Extracts

Introduction and Background

During 2005 approximately 2,500 MW of wind capacity was added in the United States, which brought installed wind capacity to about 9,150 MW. Although the total wind capacity in the United States is less than in some countries, wind energy has caught the attention of some utilities that depend on natural gas to generate power. There is evidence that wind development will continue at significant levels in the United States for the next several years, although it may be sensitive to a number of factors that include transmission availability, wind turbine availability, prices of wind technology and competing fuels, production tax credit availability, and states’ renewable portfolio standards (RPSs).

This trend has helped induce electricity providers to investigate the potential impact of wind on the power system. Because of wind power’s unique characteristics, many concerns are based on the increased variability that wind contributes to the grid, and most U.S. studies have focused on this aspect of wind generation. Grid operators are also concerned about the ability to predict wind generation over several time scales.

In this report we discuss some recent studies that have occurred in the United States since our previous work [2, 3]. The key objectives of these studies were to quantify the physical impacts and costs of wind generation on grid operations and the associated costs. Examples of these costs are (a) committing unneeded generation, (b) allocating more load-following capability to account for wind variability, and (c) allocating more regulation capacity. These are referred to as “ancillary service” costs, and are based on the physical system and operating characteristics and procedures. This topic is covered in more detail by Zavadil et al. [4].

Time Frames of Wind’s Impact

Wind can have an impact on several time scales that correspond to grid operations. The shortest is generally in the range of milliseconds to seconds, and is the domain of system dynamic stability studies. Most wind integration studies focus on longer time scales, but the stability time frame is a concern and recent developments in the United States have addressed this issue. The most important is the Federal Energy Regulatory Commission (FERC) limited grid code for wind plants, contained in FERC Order 661A, issued in December 2005 [5]. This ruling addresses the issues of low-voltage ride-through, reactive power supply, and Supervisory Control and Data Acquisition (SCADA) system requirements.

Figure 1 illustrates the key time frames that correspond to utility/grid operations and that have been the focus of most integration studies. In the United States, the regulation time frame is the period during which generation automatically responds to deviations in load or load net wind. This capability is typically provided via automatic generation control, and is a capacity service generally covering seconds to several minutes. Integrating wind into the system would have an impact on regulation requirements for the system, and might require additional regulation capability. In the United States there are two controlled performance standards, CPS-1 and CPS-2*, control area operators/balancing authorities follow.......

The second time frame is load following. This is a longer period during which generating units are moved to different set points of capacity, subject to various operational and cost constraints. Load following involves capacity and energy, and corresponds to time scales that may range from 10 minutes to a few hours. Loads can typically be forecast with reasonable accuracy and overall correlation between individual loads tends to be high in this time frame. Generating units that have been previously committed, or can be started quickly, can provide this service, subject to physical constraints. Beyond the maximum and minimum generation constraints, the ramping constraint (ability to move in MW/minute) may be affected by significant wind generation on the system. In systems with little or no wind, the changes in load in this time frame can be predicted with varying degrees of accuracy. To the extent that forecasts are wrong, the system operator must deal with the resulting system imbalance. Significant wind capacity can increase the uncertainty and cost in this time frame.

Planning for the required quantity of generation and load following capability involves the unit commitment time frame, which can range from several hours to a few days. Scheduling too much generation can increase costs needlessly, whereas insufficient generation could have a cost component (buying at high market prices or running expensive quick-start units) and a reliability component (if sufficient generation has not been started and is not available on short notice).

Most of the studies described here estimate the increased cost of managing a system with significant wind generation. The studies approach the cost question by starting with the physical behavior of the system without wind, then detailing how that physical behavior is affected by wind power plants. The primary objective of the studies is to take the view of the system operator, whose goal is to obtain system balance within required limits. Although U.S. terminology differs somewhat from that in Europe, the key physical issues and time frames are very similar. The imbalance impacts of wind are seen as unscheduled interchanges or frequency changes on the system when the balancing area cannot respond quickly enough to changes in load or wind. The impacts of wind on conventional generation are best analyzed over several time scales that correspond to system operation, ranging from automatic response (regulation in the United States) of units on automatic generation control, to spinning or standing reserve response (load-following in the United States). From the control room, wind variability is combined with load variability over these time scales, along with unscheduled deviations from some conventional generators. This net load is seen by the operator and must be balanced. Although the analytical tools differ somewhat, several common elements in the analyses have taken place in the United States.

Most of the studies we summarize here are cost-of-service studies that examine the cost of wind in the context of regulated utilities. Other studies, such as the one carried out in New York (discussed below) are market studies that do not directly calculate cost impacts. Because of this approach, the results of the market-based studies cannot be directly compared with ancillary cost studies.

Conclusions and Insights

Given the work that has been done, several conclusions are emerging. Although wind imposes additional operating costs on the system, these costs are moderate at penetrations expected over the next 5-10 years. These results are expected to apply as additional wind generation is developed in the next few years in response to state government RPSs, although wind integration costs will increase with penetration.

Large, diverse balancing areas with robust transmission tend to reduce wind’s impact and ancillary service cost. At current U.S. levels, the impact on regulation and load following appear to be modest, and the unit commitment time scale appears to be more important. In this time scale wind forecasts can play a more prominent role, and improvements in forecasting technology will certainly mitigate wind’s integration costs. As wind penetration increases in the United States, better forecasting is expected to play a more important role. To be effective, forecasts do not need to be perfect, although increasing accuracy tends to reduce costs. It is possible that at some point the incremental cost of forecast improvements will outweigh the incremental benefits that accrue from increased accuracy.

Aside from large balancing areas, other factors can mitigate wind impacts. If several adjacent balancing areas can develop cooperative arrangements or markets for ancillary services, larger quantities of wind could be absorbed because of the greater load and wind diversity that would be expected across broader regions. This could be captured by larger balancing areas, but other means of tapping this potential can be used. This is discussed further by Kirby and Milligan [7].

There is also some evidence that system operators will become more familiar with wind after working with it. For example, The Western Farmers Electric Cooperative (WFEC) in Oklahoma recently performed an analysis with the National Renewable Energy Laboratory of the operational impact of wind on its system. WFEC has a peak load of 1,400 MW and installed wind capacity of 74 MW. Initially the system operators could not maintain the CPS-1 frequency standard at its pre-wind level. With experience they became familiar with the wind system and brought CPS1 into its pre-wind range [13].

Emerging Best Practices and Methods

Although there are differences between studies, there appears to be some convergence on techniques and methods used to analyze wind’s ancillary service impacts. A key point is to recognize that the entire system—not individual loads or generators—need to be balanced. In the United States, this balance does not need to be perfect, but is required to fall within the statistical limits defined by CPS-1 and CPS-2. The implication for wind integration is profound: not every movement in wind generation needs to be matched one-to-one by another conventional generator.

The approaches used in recent studies generally capture the important system characteristics through detailed modeling of the relevant grid and operational practices. These representations of the system can then be simulated in a chronological environment that can observe the detailed constraints on the system that are imposed by loads and generators.

Because wind impacts occur throughout the time domain, the coincidence of loads and wind generation must be captured. Because wind speed and wind generation data are often difficult or impossible to obtain for desired time periods, an emerging approach is to construct the wind data from detailed time-calibrated mesoscale meteorological modeling for the desired time period. Normally this is accomplished by selecting load data for the study period based on recent historical data. Wind data sets can then be constructed to match the historical load period. And because wind impacts on some longer time scales may differ from year to year, the best approach is to extract multiple years of wind data that correspond to the loads in a multiyear study period, and complete several years of detailed simulations. This picks up any correlation (which may be highly nonlinear with significant phase shifts) between wind and load, and improves confidence that the results are meaningful.

Detailed meteorological modeling also allows the geographic impacts of wind to be represented as the turbines are spread over small areas (within a wind plant) or large areas (several wind plants) and picks up the impact of prevailing weather patterns that drive the wind generation and influences load.

The short-term behavior of wind power plants has been quantified by Wan [8]. The data sets indicate that wind power variability is quite low at fast time frames, and increases at progressively longer time frames. As a practical matter, this implies that wind’s impacts will be relatively small in the regulation time scale, increase at the load following time scale, and become even more significant at the unit commitment/scheduling time scale. The U.S. studies broadly support this conclusion, and as more wind operating data become available, a more realistic representation of wind in the analytic models can be captured so the results are more accurate.

Within the modeling frameworks used in the U.S. studies, the variability of wind generation is added to the already considerable variation in load. The analytic tools approximate the view of the system as seen by the operator. This implies that the statistical treatment of the wind and load time series is important and provides a realistic representation of wind’s impact on the regulation and load following time frames.

To better understand the role of forecasting, some studies have constructed wind forecasts and run the analysis with and without the forecasts. Clearly forecasting can play an important role in mitigating wind’s impacts on system operations and costs, but only if the forecast is used appropriately in the control room.