How to predict wind energy output

As the wind power industry expands, the ability to accurately predict the output from wind farms is becoming increasingly important.

Not only are wind forecasting tools essential to help ensure a wind farm runs efficiently, they are also vital during the development stage to help firms better determine the economic viability of projects before they start looking for investment.

For wind farm operators, robust forecasting models are a key way of handling variability and uncertainty in power generation and are essential for improving operations and maintenance planning, while also reducing health and safety risks. In addition, they play an important role in enabling operators to effectively schedule and sell power into the electricity market, helping energy companies to manage their generation portfolio and balance output on a regional or national scale.

Sound models allow energy analysts and traders to produce more accurate wind power forecasts and predict electricity prices, leading to more informed strategic decisions on the trading markets. For investors, better forecasting brings higher revenues, and is an important source of competitive advantage for both debt and equity investors who have previously invested in underperforming wind assets.

Weaknesses in current approaches

Operators can currently choose from a number of different forecast providers, and the accuracy of wind power forecasts have improved steadily over the last decade. Most models use the same basic framework, ie a combination of physical and statistical models, to predict the wind power output over a set period. However, whilst each of the available models is being continually refined, there is still room for improvement.

"No weather or power model is 100 per cent perfect – you can only try to minimise the errors," says Lesley Robertson, marketing manager at the UK Met Office.

In broad terms, a key weakness of many forecasting models is their inability to forecast ramping events (ie rapid changes in the output of wind power farms). There is also a need to develop better estimates of forecast uncertainty and to analyse how forecasts can be tailored towards the specific needs of various users.

On the technical side, a key limitation is that many operators currently use rotating cup anemometers and wind-vanes placed on the back of the turbine behind the rotor blades to measure the wind and control their turbines. The problem is that this instrumentation often generates inaccurate measurements, using turbulent wind, which is disturbed as it passes through the turbine blades, to try and judge wind speed and direction.

"It's like driving a car by looking through the rear view mirror: it's far from accurate in determining where the wind will go next and makes turbines less efficient and more expensive to operate and maintain," argues Phil Rogers, president and chief executive of Virginia-based Catch the Wind.

Looking ahead - emerging models and technology

To help improve the predictability of wind energy output, several organisations have made great strides in improving forecasting models.

In Europe, ANEMOS (an acronym for the rather unwieldy Development of A Next-Generation Wind Resource Forecasting SysteM for the Large-Scale Integration of Onshore and OffShore Wind Farms project) is a EU research and development project. Its main aims are to develop accurate statistical and physical wind forecasting models that considerably outperform the existing state of the art, for onshore and offshore wind resource forecasting, with an emphasis placed on integrating high-resolution meteorological forecasts. In the case of offshore wind, marine meteorology will be employed, as well as information taken from satellite radar images.

In the UK, the Met Office has developed an alternative system known as VisualEyes, a web-based system for managing site conditions at wind farms. The system uses Numerical Weather Prediction models to produce weather forecasts, taking thousands of observations from sites around the world. It can then produce site-specific forecasts, as well as probabilistic forecasts describing the likelihood of certain weather conditions occurring.

A web-based map viewer also allows operators to monitor conditions like lightning, wind speed, rain, temperature, snow and visibility across multiple sites, and alerts them when predefined thresholds are crossed.

Similarly, Previento is a German-designed operational system developed by the energy meteorology research group at the Carl von Ossietzky Universität Oldenburg and ForWind, the centre for wind energy research based in Oldenburg. The system predicts the power output of wind farms and allows energy providers, energy brokers, and grid operators to integrate the fluctuating input of wind energy into their daily operations. It aims to provide a reliable prediction of the expected wind power up to four days in advance for any selected location in Germany and Europe.

Forecasts are based upon local conditions in the vicinity of the wind farms, as well as on a numerical weather prediction. The entire input of wind-generated electric current from a region is calculated on the basis of selected wind farms. During this process, the representative areas are selected so that they closely mirror the regional distribution of wind energy facilities.

In contrast to other prediction systems, Previento also calculates the margin of error in relation to the actual prediction value dependent upon the current uncertainty in weather conditions.

Progress is also being made on the other side of the Atlantic where the Argonne National Laboratory (ANL) near Chicago is working with INESC Porto to conduct research on forecasting and its use in electricity markets. The consortium is working on improving the statistical models used to generate the best possible forecast from all available information. On the operational side, it is also investigating how to incorporate forecast uncertainty into the procedures used by operators throughout the power system.

"Consistent and representative uncertainty information is of high importance for many operational decisions, such as scheduling, trading, and dispatch," says Dr Audun Botterud of ANL.

Also in the US, Catch the Wind has developed the Vindicator, a laser wind-sensor that could represent the future of wind forecasting.

The sensor is mounted on a wind turbine to determine wind speed and direction, and orient the turbine into the wind before it arrives. This enables turbines to "see" the wind in 3D before it reaches a turbine's blades and realign for optimal power output.

The system was recently successfully tested on the WindSentinel, a wind resource assessment buoy developed to assist offshore developers in determining the available wind resource at potential wind-farm sites.

The Vindicator sensor emits laser beams which reflect back from airborne particles to measure wind speed and direction, gusts and turbulence, at ranges up to 300 metres. Data is run through computer algorithms to measure wind speed and direction before it reaches the turbine, giving the control system time to proactively adjust to the approaching wind.

"We believe our approach will change the way wind-farm operators measure the wind and, in doing so, allow them to generate more power by operating more intelligently and efficiently," says Catch the Wind's Rogers.

In summary, although the process of wind forecasting continues to be hampered by inherent uncertainties, models and technological approaches like the ones outlined here can help in significantly reducing the risks associated with wind-farm development.