Going with the flow – Dr Daniel Dehtyriov’s innovative use of AI modelling powers the future of wind turbines
Dr Daniel Dehtyriov is tackling the unpredictability of fluid dynamics in wind turbines by using AI modelling to enhance the efficiency of wind farm design and the renewable power generated.
Daniel’s journey to science came about quite suddenly at around 12 years old, when he discovered the joys of abstraction. “I asked my maths teacher why we use letters in place of numbers, after all we only have 26 of them. That the world around us could be, to my surprise and delight, written down and shaped by this language, satisfied both a natural curiosity and an aesthetic pleasure”. Though he often rebelled against this idea, he did so with the knowledge that it was against his true nature and sooner or later he’d return to science.
Daniel was particularly drawn to fluid mechanics because it lies precisely at the intersection of abstraction and the natural world. He notes: “Fluids are remarkably difficult to intuit without a mathematical framework; my early twenties were characterised by a radical shift from a view of scientific inquiry as purely an aesthetic pursuit to one which shapes and is indeed itself shaped by society”. Daniel highlights that one of the biggest challenges in our time is how to sustainably power the world, and believes that scientists must rise to this challenge. Daniel concentrates his efforts on renewable energy, in particular enhancing the economic feasibility of wind-energy projects. He is most enthused by the cross disciplinary nature of the science behind renewable energy; it spans the fields of applied mathematics, engineering, physics, and computational science. He adds that we are living through a quiet revolution in how science is practiced and its possibilities; to be part of that shift, help to shape it, and to direct it towards solving real world problems is a rare and compelling privilege.
Using AI in his research has not been as much of a barrier as Daniel anticipated; he has been impressed by the widespread availability of introductory resources for AI tools. However, he notes that the real challenge has been avoiding the temptation to treat AI as a universal solution: “The key to AI in science is understanding where and which AI tools can add genuine value; and that judgement comes from domain knowledge,” he says. Daniel advises that learning about the AI tools and how they work is important, but just as vital is to read widely and engage with domain expects, in order to understand where they truly apply.
Wind turbine simulation
Daniel recognises that AI is quietly revolutionising many areas of computational science. “In my field, the most interesting pivot has been the rise of physics-constrained neural networks - the idea that we can combine the strengths of machine learning with our mathematical understanding of physical systems. This fusion of AI and physical reasoning opens the door to solving longstanding challenges and engineering new solutions in renewable energy science.” Though the use of AI has opened many doors for Daniel’s research, the sparsity of high-quality training data has been a challenge when applying AI to the multiscale differential equations that govern fluid flow in renewable energy systems. His work has proposed powerful online training methods that learn from the data generated during the simulations themselves – this approach enables accurate modelling of complex, multiscale dynamics, without requiring large pre-existing data sets. The belief is that these techniques will have broad potential to support advances in wind farm control, numerical weather prediction, and fusion energy.
A defining moment for Daniel was realising that AI could not just be used as a black-box tool, whose output doesn’t necessarily satisfy our well-established physical understanding, but as a way to augment and accelerate physics-based models. This changed the trajectory of his research from solving traditional fluid dynamics problems to developing methods that fuse machine learning with the governing equations themselves. For Daniel, it has opened up entirely new ways to think about modelling, simulation, and control in renewable energy systems. The Schmidt AI in Science Fellowship programme translated this realisation into practical expertise, setting the trajectory for his future career.
