Page 48 - North American Clean Energy July/August 2019 Issue

48
JULY•AUGUST2019 /// www.nacleanenergy.com
wind power
The Wind Industry
Wakes Up to Arti cial
Intelligence
by Glen Wagner and Jim Kiles
Whether it’s improving the accuracy of a medical diagnosis, boosting student outcomes, or advancing autonomous vehicle capabilities, arti cial intelligence (AI) has become embedded into many aspects of our lives. Now you can add wind energy to the ever-growing list of industries that are capitalizing on AI and machine learning to drive greater operational e ciencies.
While interest and investment in wind energy has continued to grow (as utilities seek to diversify their energy portfolios to include more clean energy assets) it’s hardly a new technology. Wind energy has been used for thousands of years in regions around the world. Although today’s wind farms utilize sophisticated technology, the underlying methodologies for wind farms have largely stayed the same. As such, in most cases, wind energy producers are not able to maximize their aggregate power output. By combining AI, machine learning, cloud-based computing, advanced power applications, and networking capabilities, however, it is possible to optimize for a healthy boost in energy production - with commensurate Power Purchase Agreement (PPA) or high-demand-period gains. How healthy a boost? e predicted increase in energy output for optimized wind farms ranges from 2-10 percent.
Using Optimization to Boost Megawatt Output
Reducing wind wake, or wake turbulence, is the underlying principle behind integrated wind energy optimization. In traditional wind farms, each turbine is individually optimized. Much like an airplane
or boat, each turbine creates its own wake. In
this scenario, the wake from the upwind turbines prevents downwind turbines from receiving full energy from the wind stream. is is not an isolated problem: It is estimated that two-thirds of U.S. wind farms experience reduced capacity due to wind wake.
Arti cial intelligence enables wake steering, whereby upwind turbines steer the wake away from downwind turbines by adjusting the yaw (or side-to-side) settings of each turbine. In doing so, the turbines can work cooperatively to extract maximum energy from the wind stream.
Wind physics and machine learning algorithms are used to perform complex calculations
that determine the optimal yaw settings for
all conceivable data combinations, including historical weather and wind patterns, elevation, types of turbines (including thrust and power curves), exact turbine locations and other factors. Specially architected control technology uses a fault-tolerant network to communicate data from each turbine’s wind sensor to the optimization platform every 1-2 seconds. is near-real-time data enables a more accurate calculation of the optimal yaw setting for each turbine within a wind farm. Machine learning is then applied to the data generated from all turbines on the network to further improve optimization models over time.