A simple software upgrade could improve the efficiency of wind turbines by ensuring they face directly into the wind more of the time. If the technique were rolled out worldwide, it could boost electricity production by 5 terawatt-hours a year 鈥� about the same amount as is consumed annually by Albania or 1.7 million .
Wind turbines work most efficiently when directly facing oncoming wind, but changing wind patterns mean they must regularly change direction around their tower. The more time a turbine spends facing even fractionally the wrong way, the less efficient it is.
Current algorithms track wind patterns and adjust the turbine blades to anticipate changes, but and at the Polytechnic Institute of Paris believe artificial intelligence can do better. They have trained a reinforcement-learning algorithm to monitor wind patterns and develop its own strategy to maintain the turbine at the correct angle. And because reorientating the blades causes wear and tear on components and consumes energy, they also tasked the AI to restrict these moves to less than 10 per cent of the time.
Advertisement
The researchers ran computer simulations testing the same wind conditions on wind turbines running different control algorithms. They found that the AI aligned their simulated version of a REpower MM82 2-megawatt turbine more efficiently than the other algorithm, resulting in the average number of degrees off the optimum angle falling by 11.2 per cent.
The AI moved the turbine more often than the other algorithm, but saved power and kept the time spent readjusting to only 3.7 per cent. The AI model produced power gains of 0.4 per cent, and still offered an advantage of 0.3 per cent once the extra energy spent rotating the turbine was taken into account.
If applied to all wind turbines in the world, this would equate to boosting their combined electricity production by 5 terawatt-hours annually, say the researchers.
at the University of Exeter, UK, says wind farms already typically operate at more than 90 per cent efficiency, but it is feasible that a software update could bring improvements.
Cochrane says companies are keen to eke out greater productivity from wind turbines. Vertical-scanning lidar devices mounted on turbines already use laser pulses to track particles of dust moving through the atmosphere to anticipate wind direction changes, and horizontal lidar sensors fitted will soon give even more warning, says Cochrane. But he believes the systems aren鈥檛 mutually exclusive.
鈥淢aybe the AI can work with the lidar to optimise the adjustment,鈥� says Cochrane. 鈥淭here might be tiny calibration errors on the weather vane and your AI could correct for that.鈥�
at Durham University, UK, says that a 0.3 per cent improvement in output can represent a large boost in an industry where investment and revenue can both be huge.
鈥淭he 0.3 per cent is important. Wind farm companies are interested in chasing the 0.1 per cents, the 0.3 per cents. It could be the difference between them having the revenue to develop the next one or not,鈥� he says.
arXiv
