AI-Driven Optimization Boosts Hydropower Turbine Efficiency in DLR and Voith Hydro Collaboration

DLR and Voith Hydro Show Early Successes in AI-Driven Turbine Development

Voith Hydro, a global leader in modern hydropower plants, has partnered with the DLR Institute of Propulsion Technology to improve water turbine design using AI-assisted optimization. This collaboration focuses on enhancing turbine development by applying intelligent optimization methods that boost efficiency and deepen insight into complex flow behavior.

Hydropower remains a key renewable energy source. However, modern turbines face growing demands for maximum efficiency, durability, and operational safety amid tighter economic and regulatory requirements. Designing complex flow components like runners and guide vanes poses significant challenges. Meanwhile, digital, data-driven processes are becoming critical to staying competitive and reducing development time and costs. New design approaches are essential to meet these pressures.

The DLR Tool AutoOpti: Optimizing Complex Turbine Systems

At the core of this effort is AutoOpti, an optimization tool developed by DLR for automated, interdisciplinary design and optimization processes. It integrates fluid dynamics, structural mechanics, and other simulation models with intelligent strategies for design selection and model reduction. Together with Voith Hydro, AutoOpti is applied to flow-mechanical optimization of turbine components such as rotors and guide vanes.

This enables targeted adjustments to minimize hydraulic losses and mechanical stress far faster than traditional methods.

AutoOpti Features for Industrial Optimization

• AI-based multi-objective optimization handling high-dimensional search spaces with surrogate models, variable simulation fidelities, and gradient data.
• Parallelized for high-performance computing (HPC) environments.
• Customizable quality criteria to balance optimization precision and runtime.
• Proven in aerospace, energy, and propulsion industries.
• Open interfaces supporting diverse simulation models and easy integration with existing CFD and CSM software.

Advancing Turbine Development with Machine Learning

The project incorporates real geometries and boundary conditions from Voith’s product development into a fully automated optimization cycle. AutoOpti generates new design variants, runs simulations autonomously, and evaluates results based on targets like hydraulic efficiency, flow losses, and mechanical robustness.

Using Gaussian Process surrogate models reduces the number of expensive CFD calculations required. This combination allows systematic and efficient exploration of the design space.

Initial Results: Clear Efficiency Gains

Early tests on a Francis turbine showed that AutoOpti produced new designs with notable efficiency improvements across multiple operating points while keeping flow rates constant. Surrogate models trained on high-fidelity CFD data integrated smoothly into the process.

These outcomes demonstrate that AutoOpti shortens development cycles and improves turbine design quality.

Optimization Down to the Last Detail: AI-Enhanced Runner Design

A side-by-side comparison shows the difference between a conventionally designed turbine runner and one optimized via AutoOpti, highlighting the tool’s ability to refine critical components effectively.

Outlook: Expanding Applications and Ongoing Development

The next steps include extending AutoOpti to additional turbine components and incorporating more physical disciplines. Voith is also exploring applying AutoOpti’s optimization methods to other product lines.

The DLR Institute of Propulsion Technology supports this with workshops, training sessions, and integration assistance to embed these methods into existing workflows. The long-term goal is to establish this collaboration as a lasting development partnership.

Beyond current Gaussian Process models, the institute is developing new deep neural network approaches using transformer-based architectures called FlowFormer, which aim to directly approximate 3D flow behavior.

For professionals in product development interested in AI-assisted optimization and simulation, exploring advanced tools like AutoOpti offers valuable insights into accelerating design cycles and improving product performance.