Wind turbines fight. Did you know that? In wind farms - like large offshore arrays - turbines compete for power, with more kinetic energy being extracted from the air by the upstream turbines than by downstream ones. The turbines are coupled by their aerodynamics, as well as being coupled by their power electronics network.
There are three upshots of this:
- Large amounts of power conditioning are required, to overcome highly nonlinear interaction effects.
- Turbine control systems are run conservatively, working on their own to provide target outputs, instead of together as a larger network. This limits production.
To address this, work is progressing on modular "Digital Twins" - where pieces of hardware - like inverters - are replicated in a real-time simulation of an asset. A digital twin of an array can comprise an assembly of sub twins, each with its own AI and/or simulation code. Twins can be used as a tool to train AIs in how to control plant.
The goal of the project (which is ongoing) is to train an AI to optimise the control state of each turbine for net benefit as a whole array (longer lifetime, increased production, etc).
This should give insight into how an AI can be used to manage nonlinear aerodynamic interactions between turbines and how to increase the production.
But, that does nothing about the power network interactions. It would be super cool to:
The goal of adding an SDI would be to work toward a proof-of-concept for use of SDIs in wind farm applications - by removing the network coupling effect between the turbines, an AI farm controller of the type we're working on would have much greater scope to improve overall array production.
Progress with the design of the model wind turbines (with blade pitch mechanism for control) is well under way and prototypes have been printed. Generators (tiny ones!) have been selected. Images are attached showing the main driveshaft and blade pitch assemblies.
The general arrangement will be based on the NREL 5MW wind turbine. We'll be using a ducting system to ensure that we get some aerodynamic coupling between model turbines... at this really small scale it won't be identical aerodynamics but it will at least represent an interaction.
We get that TAPAS boards are overkill for this project in terms of the power they can handle (!!) - but they could provide a perfect platform for representing the power conditioning electronics.
Tom Clark and Wolfgang Schmied.
Tom is the Cofounder of Octue.com, a platform for building digital twins of engineering assets (this project is not affiliated with Octue, although uses some of the industry knowledge and experience that Tom has gleaned over 10 years in the wind industry)
Wolfgang is a bioscientist and 3d printing enthusiast, a previous winner of the bio-maker challenge, with "A low-cost, pressurized liquid chromatography system for protein purification".
Number of boards
Is it too much to ask for 6? That's one for each turbine? We could still use one to invert the output of the whole farm (simple demo) but if we had 6 we couldl attempt to show how they can be used to eliminate the nonlinear fighting between the turbines.
this idea gets sixTAPAS dev kits