Experience in the Duke Energy Carolinas service territory shows that high penetration rates for electric vehicles, combined with increased natural gas-fired power generation, can result in lower...
Leaning on Line Pack
Green energy mandates might overburden gas pipelines.
to construct and operate offshore wind farms. One inconvenient characteristic of onshore wind production is its relatively greater variability, compared to offshore wind. Associated with this greater variability is much higher prediction error in the day-ahead (DA) and hour-ahead (HA) forecasts of wind production. The DA wind forecast in the latest wind integration study performed in New England shows an overall forecast accuracy of 15 percent to 20 percent mean absolute error (MAE). Forecast accuracy of 15 percent to 20 percent MAE is considered state-of-the-art. 5 By comparison, the load forecast error is typically 2.5 percent in peak months and less than 1.5 percent in off-peak months.
Wind power forecasting (WPF) involves the use of complex stochastic or probabilistic models that draw upon weather prediction results, local meteorological measurements, terrain and topography details, and supervisory control and data acquisition (SCADA) data from the wind farms. There are many different WPF models for the ISOs to choose from. Some ISOs have even done pilot studies with different models in order to identify the best model for their distinctive weather patterns and terrains. The performance of the models is strongly linked to the terrain complexity of the region, such that in one benchmarking study the average value of the normalized MAE ranged between 10 percent for flat terrain to 21 percent for highly complex terrain. 6 The WPF error is highly dependent on the wind speed forecast error, which itself depends largely on the numerical weather prediction global model. Forecast accuracy can be improved by using a combination of different forecasts, either from different WPF models or different numerical weather prediction models.
To offset the significant prediction error that is inevitably part of forecasting DA and HA wind production, ISOs are expected to rely increasingly on ancillary services provided by spinning and non-spinning reserves. The NYISO study on increased wind generation found that system variability increases and varies by season, month, and time of day, leading to higher magnitude ramping. 7 Higher ramping requirements are tantamount to greater changes in net load over time, to which the dispatchable resources need to respond. Holding constant existing resource adequacy and operational reliability criteria, only about 0.2 MW to 0.3 MW of existing conventional resources can be retired with the addition of 1 MW of wind. 8 The California ISO has found that the addition of solar resources can lessen operational requirements in some hours but increase them in others, compared to wind generation alone. 9
Fundamental weaknesses in regional capacity markets—that is, low clearing prices due to capacity overhangs, the economy and the ascent of demand response (DR)—make it increasingly difficult for old-style steam turbine generators to remain in the market. This is particularly true for those facing significant capital outlays for environmental compliance.
Also, among the recent fleet of combined cycle (CC) plants, there have been bankruptcies and recapitalizations as assets have changed hands at bargain-basement prices. As gas plants lose market share in terms of energy sales to inframarginal wind plants, the supply of ancillary services available from quick-start and higher magnitude ramping plants