Intermittency is a major problem that may well limit the penetration of wind power generation. The 2006 Energy in Scotland Inquiry report expresses concern about some aspects of wind power.

"The inherent intermittency of wind power means that it cannot be relied on to deliver ?rm output at any given time. However, its input when available has to be accepted into the grid. A diversity of supply is essential to achieve maximum security and ?exibility in the supply of electricity."

A study commissioned by the US state of Minnesota considered penetration of up to 25%, and concluded that integration issues would be manageable and have incremental costs of less than one-half cent ($0.0045) per kWh, while a similar report from Denmark noted that their wind power network was without power for 54 days during 2002.

Since wind speed is not constant, a wind generator's annual energy production is never as much as its nameplate rating multiplied by the total hours in a year. The ratio of actual productivity in a year to this theoretical maximum is called the capacity factor. A well-sited wind generator will have a capacity factor of about 35%. This is due to the variable nature of wind. Capacity factors of other types of power are based mostly on economics, with a small amount of downtime for maintenance. Nuclear plants have low fuel cost, and are therefore often run constantly at full output, with the load following relegated to other plants, and thus typically achieve a 90% capacity factor.^[8] The lower values of 70% for coal plants and 30% for oil plants reflect a throttling-back of plants with high cost fuel in times of low demand. According to a 2007 Stanford University study published in the Journal of Applied Meteorology and Climatology, interconnecting 10 or more well-sited wind farms over a dispersed geographic area allows roughly 1/3 of the total energy produced to be relied on for baseline loads.

Storage, such as with pumped hydroelectric storage, can be used to "shape" wind power (by assuring constant delivery reliability), adds a cost of about 25% to yield viable commercial performance. Electricity consumption can be adapted to production variability to some extent with Energy Demand Management and smart meters that offer variable market pricing over the course of the day. For example, municipal water pumps that feed a water tower do not need to operate continuously and can be restricted to times when electricity is plentiful and cheap. Consumers could choose when to run the dishwasher or charge an electric vehicle, making it very convenient. Electric and plug-in hybrid vehicles also offer a significant demand management tool and could potentially be set to charge automatically during periods of excess wind output.

2.1 Distribution of wind speed

Windiness varies, and an average value for a given location does not alone indicate the amount of energy a wind turbine could produce there. To assess the climatology of wind speeds at a particular location, a probability distribution function is often fit to the observed data. Different locations will have different wind speed distributions. The distribution model most frequently used to model wind speed climatology is a two-parameter Weibull distribution because it is able to conform to a wide variety of distribution shapes, from Gaussian to exponential. The Rayleigh model, an example of which is shown plotted against an actual measured dataset, is a specific form of the Weibull function in which the shape parameter equals 2, and very closely mirrors the actual distribution of hourly wind speeds at many locations.

Worldwide installed capacity and prediction 1997-2010, Source: WWEA

Because so much power is generated by higher wind speed, much of the average power available to a windmill comes in short bursts. The 2002 Lee Ranch sample is telling; half of the energy available arrived in just 15% of the operating time. The consequence is that wind energy does not have as consistent an output as fuel-fired power plants; utilities that use wind power must provide backup generation or grid power reception capability for times that the wind is weak.

Electricity generated from wind power can be highly variable at several different timescales: from hour to hour, daily, and seasonally. Annual variation also exists, but is not as significant. This variability can present substantial challenges to incorporating large amounts of wind power into a grid system, since to maintain grid stability, energy supply and demand must remain in balance.

While the negative effects of intermittency have to be considered in the economics of power generation, wind is unlikely to suffer momentary failure of large amounts of generation, which may be a concern with some traditional power plants. In this sense, it may be more reliable (albeit variable) due to the distributed nature of generation. That said, winds often stagnate during periods of peak demand, such as during heat waves.

Wind speeds are generally much lower during periods of the highest peak-load demand (the months of June, July and August) in North America. There is an inverse relationship with wind speed and peak demand of electricity. Many grid planners do not even adjust their calculations to account for wind power installations because of that inverse (albeit happenstance) relationship.

2.2 Grid management

Grid operators routinely control the supply of electricity by cycling generating plants on or off at different timescales. Most grids also have some degree of control over demand, through either demand management or load shedding. Management of either supply or demand has economic implications for suppliers, consumers and grid operators but is already widespread.

Variability of wind output creates a challenge to integrating high levels of wind into energy grids based on existing operating procedures. Critics of wind energy argue that methods to manage variability increase the total cost of wind energy production substantially at high levels of penetration, while supporters note that tools to manage variable energy sources already exist and are economical, given the other advantages of wind energy. Supporters note that the variability of the grid due to the failure of power stations themselves, or the sudden change of loads, exceeds the likely rate of change of even very large wind power penetrations.

There is no generally accepted "maximum" level of wind penetration, and practical limitations will depend on the configuration of existing generating plants, pricing mechanisms, capacity for storage or demand management, and other factors.

A number of studies for various locations have indicated that at least 20% of the total electrical energy consumption may be incorporated with minimal difficulty. These studies have generally been for locations with reasonable geographic diversity of wind; suitable generation profile (such as some degree of dispatchable energy and particularly hydropower with storage capacity); existing or contemplated demand management; and interconnection/links into a larger grid area allowing for import and export of electricity when needed. Beyond this level, there are few technical reasons why more wind power could not be incorporated, but the economic implications become more significant and other solutions may be preferred.

At present, very few locations have penetration of wind energy above 5%. Germany, Spain, and Portugal all have penetration levels below 10%, however, and Denmark's penetration is over 20%, demonstrating that the technical issues are manageable at relatively high levels. The penetration of intermittent power sources in Denmark is even higher since 20% of Denmark's electricity is produced by decentralize combined heat-power plants that only produce electricity when there is a demand for heat. However, it should also be noted that the Danish grid is heavily interconnected to the German and broader European electrical grid and can both supply and demand electricity from a broader area than just the Danish grid. In practice Denmark has solved its grid management problems by exporting almost half of its wind power to Norway. The correlation between electricity export and wind power production is very strong.

Induction generators typically used for wind power projects require reactive power for excitation, so typically substations used in wind-power collection systems include substantial capacitor banks for power factor correction. Groups of induction generators behave differently during transmission grid disturbances, so extensive modeling of the dynamic electromechanical characteristics of a new wind farm is required by transmission grid operators to ensure predictable stable behavior during system faults. In particular, induction generators cannot support the system voltage during faults, unlike steam or hydro turbine-driven synchronous generators.

2.3 Grid energy storage

A grid energy storage system is a potential means of increasing the amount of usable energy in a given electrical system (penetration rates) by making use of 'energy storage systems'. Effectively, "surplus" energy could be used to store electricity in usable form. Storage of electricity would effectively arbitrage between the cost of electricity at periods of high supply and low demand, and the higher cost at periods of high demand and low supply. The potential revenue from this arbitrage must be balanced against the installation cost of storage facilities and efficiency losses. Many potential technologies exist to store usable electric energy, including pumped storage hydroelectricity, air ballast also known as compressed air energy storage, battery technologies, production of hydrogen using electrolysis, and even flywheel energy storage.

2.4 Predictability

Related to, but essentially different from variability, is the short-term (hourly or daily) predictability of wind plant output. Like other electricity sources, wind energy must be "scheduled" - this presents a challenge because the nature of this energy source makes it inherently variable over time. To overcome this problem, wind power forecasting methods are employed by utilities or system operators. Despite the use of forecasting, the predictability of wind plant output remains low for a variety of reasons.