New method boosts wind power generation without new equipment

Virtually all wind turbines, which generate more than 5 percent of the world’s electricity, are controlled as if they were stand-alone units. In fact, the vast majority of them are part of larger wind farms with dozens or even hundreds of turbines whose traces can influence each other.

Now engineers at the Massachusetts Institute of Technology and others have discovered that without any new investment in equipment, the capacity of such wind farms can be increased by modeling the wind flow of the entire set of turbines and optimizing the control of individual units. respectively.

The increase in energy production from this installation may seem modest – in general, it is about 1.2%, and at optimal wind speeds – 3%. But the algorithm can be deployed at any wind farm, and the number of wind farms is growing rapidly to meet accelerated climate targets. If this 1.2 percent energy increase were applied to all of the world’s existing wind farms, it would be the equivalent of adding more than 3,600 new wind turbines, or enough to power about 3 million homes, and the total gain for electricity producers would be nearly a billion dollars. dollars a year, the researchers say. And all this is practically free.

The study was published today in the journal energy of nature, in a study led by MIT’s Esther and Harold E. Edgerton, Michael F. Howland Associate Professor of Civil and Environmental Engineering.

“Essentially, all existing utility-scale turbines are driven greedily and independently,” says Howland. The term “greedily,” he explains, refers to the fact that they are controlled to maximize only their own energy production, as if they were isolated units with no detrimental effect on neighboring turbines.

But in the real world, turbines in wind farms are deliberately placed close together to capture economic benefits associated with land use (onshore or offshore) and infrastructure such as access roads and power lines. This proximity means that the turbines are often heavily affected by wakes created by other turbines opposite them, a factor that is currently not taken into account by individual turbine control systems.

“In terms of flow physics, placing wind turbines close together in wind farms is often the worst thing you can do,” says Howland. “The ideal approach to maximizing total energy production would be to place them as far apart as possible,” but that would increase the associated costs.

This is where the work of Howland and his collaborators comes into play. They developed a new flow model that predicts the power output of each turbine on the farm, depending on the winds blowing into the atmosphere and the control strategy of each turbine. Although the model is based on flow physics, the model learns from wind farm operating data to reduce prediction errors and uncertainty. Without changing the physical layout of the turbines and the hardware systems of existing wind farms, they used physics-based simulations of the flow within the wind farm and the resulting power output of each turbine under various wind conditions to find the optimal orientation for each turbine at a given moment. This allows them to maximize the performance of the entire farm, not just individual turbines.

Today, each turbine constantly detects the direction and speed of the oncoming wind and uses its internal control software to adjust the position of the yaw angle (vertical axis) to get as close to the wind as possible. But in the new system, for example, the team found that by turning one turbine slightly away from its own maximum power position—perhaps 20 degrees from its individual peak power angle—the result was increased power output from one or more associated engines. . units more than compensate for the slight decrease in output compared to the first unit. Using a centralized control system that took into account all these interactions, the group of turbines operated at power output levels that were 32 percent higher in some conditions.

In a months-long experiment at a real wind farm in India, the predictive model was first tested by testing a wide range of yaw orientation strategies, most of which were intentionally sub-optimal. By testing many management strategies, including suboptimal ones, both on a real farm and on a model, the researchers were able to determine the true optimal strategy. Importantly, the model was able to predict the farm’s power generation and optimal control strategy for most of the wind conditions tested, giving confidence that the model’s predictions will track the true optimal farm strategy. This allows the model to be used to develop optimal control strategies for new wind conditions and new wind farms without having to do new calculations from scratch.

Then, a second multi-month experiment on the same farm, in which only the optimal control predictions from the model were implemented, proved that the real effects of the algorithm could match the overall improvements in energy consumption observed in the simulation. On average over the entire test period, the system achieved an energy output increase of 1.2 percent at all wind speeds and 3 percent at 6 to 8 meters per second (13 to 18 mph).

While the test was conducted at a single wind farm, the researchers say the co-management model and strategy could be implemented at any existing or future wind farm. Based on the world’s existing wind turbine fleet, Howland estimates that a total energy improvement of 1.2 percent would produce more than 31 terawatt-hours of additional electricity per year, roughly equivalent to installing an additional 3,600 wind turbines for free. This will bring wind farm operators an additional $950 million a year in revenue, he said.

The amount of energy received will vary widely from one wind farm to another, depending on many factors, including the distance between the units, the geometry of their location, and changes in the wind pattern at that place during operation. year. But in all cases, the team’s model can give a clear prediction of exactly what the potential benefits for a given site are, Howland says. “The optimal control strategy and potential energy gain will be different for each wind farm, which prompted us to develop a predictive wind farm model that can be widely used to optimize the wind farm park,” he adds.

But the new system could potentially be adopted quickly and easily, he says. “We do not need to install additional equipment. We’re really just making changes to the software, and there’s a significant potential increase in energy associated with that.” He points out that even a 1% improvement means that in a typical wind farm of around 100 units, operators can get the same power with one fewer turbine, saving costs, typically millions of dollars, associated with acquisition, construction and maintenance. . installation of this block.

In addition, he notes, by reducing wake losses, the algorithm could allow turbines to be placed more closely together in future wind farms, thereby increasing wind power density and saving onshore (or offshore). This increase in power density and reduction in footprint can help achieve the urgent goals of reducing greenhouse gas emissions, which require a significant expansion of wind power use both onshore and offshore.

What’s more, the biggest new wind farm development area is offshore, he says, and “the impact of wake loss is often much higher at offshore wind farms.” This means that the impact of this new approach to managing these wind farms could be significantly greater.

Howland Lab and the international team continue to refine the models and work to improve the operating instructions they receive from the model, moving towards autonomous co-management and aiming for the highest possible power output under a given set of conditions, Howland says. .