But how can we maximise energy yields and minimise fatigue loads, to keep costs low while ensuring that wind farms behave more like conventional power stations?
It's necessary, at least in the short-term, to smooth the path towards a future decarbonised grid system, which will ultimately look very different.
DNV GL recently published its Energy Transition Outlook, which projects the developing energy mix out to 2050.
It forecasts wind-power installation rates will more than double by 2030, double again by 2040, and yet again by 2050. Variable renewable-power sources, mainly wind and solar, will form the bulk of our electricity generation.
The weather-dependency of these power sources presents challenges for grid operators to ensure supply security and maintain short-term stability of frequency and voltage.
Wind power can already do much to present itself to the grid like a conventional power plant; and effective performance monitoring and maintenance scheduling is more important than ever to ensure energy output is optimised and meets grid obligations.
It is important to understand the limitations and capabilities of wind projects. Forecasting output is essential for efficient grid operations, from minutes to days, and is becoming more sophisticated and reliable.
It is possible to control a wind farm's power output second by second, in response to variations in wind conditions, grid frequency and voltage, as well as direct commands from grid operators.
Although the energy available in the wind is determined by meteorological factors, there is scope to control short-term wind-farm power output with considerable flexibility, and very rapid response.
To make best use of wind farms as grid-friendly power stations, it is essential to understand wake effects, whereby upstream turbines create a wind shadow for downstream turbines, reducing their energy production and exacerbating fatigue loading through increased turbulence. This is a significant challenge.
The atmospheric boundary layer, the turbine wakes and their effects on turbine power production and loading make a highly complex physical system.
Well-validated computer models are needed that capture the main effects and are not too time-consuming to run. These models can then be used to create strategies for manipulating all the individual turbine set-points in changing meteorological conditions, which requires complex optimisation; while verifying the success of such schemes requires sophisticated field experiments.
Two main forms of control action are being investigated to manipulate the wake interactions between turbines. Induction control involves lowering the power output of some turbines, reducing the strength of their wakes, so other wake-affected turbines can increase power output and lower loads.
Wake-steering control involves yawing some turbines slightly off-wind, causing the wake to move sideways to steer wakes away from downstream turbines.
Both strategies involve a trade-off whereby improvements in power output and loading at some turbines outweigh worsened performance at others. These two methods should be used in combination for optimal performance.
Given suitable simulation models, methods to achieve optimal control range from quasi-static open-loop control or "advanced sector management" — which may work satisfactorily if wind conditions are varying slowly, to dynamic closed-loop control schemes that may offer faster response and be more tolerant of imperfect modelling assumptions.
Advanced sector management uses pre-calculated tables, where operators can look up turbine set-points, depending on average meteorological conditions across the site derived from conventional measurements.
Dynamic closed-loop control interprets detailed real-time measurements across the whole wind farm and attempts to track the individual wakes, taking into account the wake propagation time delays.
Current research is aimed at developing appropriate strategies across this range, improving speed and accuracy of the underlying models, and reducing uncertainties to the point where the benefits of wind-farm control can be demonstrated commercially.
Using the latest forecasting models combined with the capability to modify output rapidly, it is already possible to operate wind farms like conventional power stations, facilitating the integration of large quantities of wind power into the electricity system in the short-term.
In the longer-term, by including rapid developments in storage and demand management, an electricity system powered mainly by renewables is within sight.
Ervin Bossanyi is a senior principal researcher for renewables in DNV GL's technology and research group