Simulations of the flow-driven rotation of the Francis-99 turbine runner. Evaluating the use of PANS and ILES turbulence modelling.

Today, the steadily increasing amount of renewable energy production is changing the way hydraulic power plants are used. The intermittent generation of wind and solar power creates instabilities in the electric grid that need to be countered in a fast and reliable way. Hydro power plants are good at changing power output by simply adjusting the flow through the turbine. They are, however, not designed for that amount of transient loads, which will result in a lower efficiency at off-design operating conditions and a higher degree of fatigue. In this thesis, the flow in a high-head Francis turbine is investigated. The turbine is a model of a full scale prototype turbine operating in Norway. It is in the spirit of the Francis-99 workshop series, aimed to determine the state of the art in high-head Francis turbine simulations, that this thesis is written. The geometry and experimental results of the Francis-99 turbine that are used in this work are freely available at the Lule˚a University of Technology homepage. The focus of the thesis is two-fold; firstly, the accuracy of a number of turbulence modelling methods are investigated at the best efficiency operating condition. The turbulence modelling approaches used are the k-ε RNG and k-ω SST URANS models, together with the k-ω SST PANS and ILES methods. Secondly, the implementation of flow driven runner rotation for the purpose of transient operation simulations is evaluated. The operating condition used to test the functionality of this implementation is a load-rejection case. All simulations are performed in an unsteady framework using the finite volume code FOAM-extend 3.1, which is a highly modifiable, open-source CFD code. The geometry comprises all parts from the spiral casing to the draft tube outlet, with the full runner and all guide vanes included. All turbulence models perform similarly when comparing the velocity profiles, with the ILES and k-ω SST models correlating best with the experiments. The strong swirl appearing in most of the simulations presented at the 2014 workshop does not appear in the present work, which is closer to the experimental results. The turbine head is over-predicted by 10-20% and the extracted shaft work by 12-21%, resulting in good agreement with experimental efficiency. The main frequencies of the pressure probes agree with experiments, and correspond to the blade passing frequencies of the runner blades and guide vanes. The implementation of the flow-driven rotation show robustness and promising trends, but further investigation is needed before comments can be made on accuracy