Periodic Reporting for period 3 - WakeOpColl (Learning and collective intelligence for optimized operations in wake flows)
Berichtszeitraum: 2020-09-01 bis 2022-02-28
WakeOpColl aims at the development of a novel bio-inspired control paradigm for groups of devices which influence each other through fluid-mediated interactions. These so-called wakes are the regions of a flow which carry the signature of an interaction with a device, because of a momentum and energy exchange. An aircraft that generates lift or a power-producing wind turbine are prototypical examples of such situations and actually the study cases of this project. A crucial aspect of wake interactions is that they can be either beneficial or detrimental for the impacted device. One might then wonder how
- an individual device could exploit a wake or alleviate wake effects
- a group of devices can collaborate to exploit these benefits in an optimal way.
These two questions actually translate to the two prototypical flows mentioned above:
1) can we elaborate collaborative control schemes for 1) aircraft formation which will would ensure radical fuel gains and 2) for wind farms to increase production and reduce the mechanical fatigue of the turbines?
2) can we use artificial intelligence in the development of smarter control at the level of a single device, which would unlock additional gains?
These issues are crucial when one considers the exponential evolution of air traffic and the related environmental impacts, and the needed growth of renewable energy.
- an individual device could exploit a wake or alleviate wake effects
- a group of devices can collaborate to exploit these benefits in an optimal way.
These two questions actually translate to the two prototypical flows mentioned above:
1) can we elaborate collaborative control schemes for 1) aircraft formation which will would ensure radical fuel gains and 2) for wind farms to increase production and reduce the mechanical fatigue of the turbines?
2) can we use artificial intelligence in the development of smarter control at the level of a single device, which would unlock additional gains?
These issues are crucial when one considers the exponential evolution of air traffic and the related environmental impacts, and the needed growth of renewable energy.
WakeOpColl researchers have made some progress on several aspects that pertain to both the modeling of wake phenomena and the actual exploitation or alleviation of wake effects.
The project researchers can now reproduce wake phenomena between aircraft or wind turbines in numerical simulations at several levels of fidelity.
Low fidelity tools can be used to train artificial intelligence algorithms guiding aircraft or wind turbines Those then "learn" to operate more efficiently in turbulence or wakes, or extract energy from the flow with a lesser impact on their mechanical structure.
These low fidelity tools, as they are computationally affordable, can be part of our AI-based controller, as they provide it with a good "mental picture" of the flow. This idea of providing a "mental picture" is applied to aircraft flying within the wake turbulence of another aircraft and to a wind turbine subjected to the wake losses of an upstream turbine.
The developed control schemes are then evaluated in computer simulations, but at higher fidelity levels, in order to assess the performance gains and the robustness of the schemes.
These efforts have led to several advances. A first one is novel wake sensing strategies for aircraft flying in formation. This sensing is crucial if one wants to safely exploit the energy benefits of flying in the upwash region of an aircraft wake.
This strategy is also being developed for wind farms, to allow wind turbines to "feel" the wakes of their upstream neighbours and help each other.
The project researchers can now reproduce wake phenomena between aircraft or wind turbines in numerical simulations at several levels of fidelity.
Low fidelity tools can be used to train artificial intelligence algorithms guiding aircraft or wind turbines Those then "learn" to operate more efficiently in turbulence or wakes, or extract energy from the flow with a lesser impact on their mechanical structure.
These low fidelity tools, as they are computationally affordable, can be part of our AI-based controller, as they provide it with a good "mental picture" of the flow. This idea of providing a "mental picture" is applied to aircraft flying within the wake turbulence of another aircraft and to a wind turbine subjected to the wake losses of an upstream turbine.
The developed control schemes are then evaluated in computer simulations, but at higher fidelity levels, in order to assess the performance gains and the robustness of the schemes.
These efforts have led to several advances. A first one is novel wake sensing strategies for aircraft flying in formation. This sensing is crucial if one wants to safely exploit the energy benefits of flying in the upwash region of an aircraft wake.
This strategy is also being developed for wind farms, to allow wind turbines to "feel" the wakes of their upstream neighbours and help each other.
The flow sensing that we have developed for aircraft and wind turbines should allow real-time tracking of wake structures within aircraft formations or wind farms.
This paves the way to the efficient control of these devices in a wake-impacted environment and actually to either alleviate the negative impacts or exploit it at one's advantage.
The next steps of the project will concern larger system of devices and the design of social-like interactions which lead the devices to collaborate and share the benefits/the burden incurred in the wake interactions.
These social interactions will have to "engineered" to let a collective behavior emerge... and produce a global optimization (of power production, fuel savings...); this naturally inspires us to tackle this bigger challenge from the perspective of game theory
This paves the way to the efficient control of these devices in a wake-impacted environment and actually to either alleviate the negative impacts or exploit it at one's advantage.
The next steps of the project will concern larger system of devices and the design of social-like interactions which lead the devices to collaborate and share the benefits/the burden incurred in the wake interactions.
These social interactions will have to "engineered" to let a collective behavior emerge... and produce a global optimization (of power production, fuel savings...); this naturally inspires us to tackle this bigger challenge from the perspective of game theory