Simulation methods for hydropower flows : modeling and validation experiments

Abstract: My research project aims for more accurate modeling of the flow in hydropower plants. By accurate modeling, it is possible to try different modifications on existing plants in an affordable manner. Also, entirely new designs can be tested before realization. Hydropower delivers almost 50% of Sweden's electricity demand with a production of 74 TWh. Hydropower plants are very efficient, but even improvements of 0.1% in efficiency corresponds to great values for the hydropower industry. Modeling hydropower plants in computer simulations gives the possibility to try new designs or modifications to existing designs that may increase the efficiency, without expensive full-scale tests. The flow in hydropower applications is very complex, putting high demands on the modeling tool. The main tool is Computational Fluid Dynamics (CFD) software. Due to the highly turbulent flow, special models have to be implemented to account for these effects. Many such models exist, but in the complex geometries and flow conditions of hydropower flows better models have to be developed. To be able to improve existing models or to develop new models, we study the flow and models in experiments and computer simulations. By generic experiments in our lab, using a water tunnel with different geometries, we can measure the real flow field and compare it to different models in simulations. In this way, we are able to isolate different phenomena in the hydropower flow (such as pressure gradients, streamline curvature, swirl, separation and unsteadiness), and to see their impact on the simulations. In this thesis (in Paper C), we present measurements in a diffuser, which is a part of the draft tube in a hydropower plant. The diffuser imposes an adverse pressure gradient on the flow, which makes modeling more difficult. The experimental equipment includes a high standard Laser Doppler Velocimetry system (Dantec) for accurate velocity measurements. Detailed measurements have been made in the near-wall region down to y+=1. This study will be a validation case for CFD simulations. The simulation part consists of evaluating existing turbulence models and developing new models. The flow region close to solid walls is an area of special interest. This is where existing models seem to fail. The near-wall region is characterized by large gradients and difficult boundary conditions for turbulent quantities. There are a number of different ways to handle this problem. The standard way, often used in industrial CFD, uses a very simplified model near the wall (wall functions). More sophisticated models use different correction functions, or even a different turbulence model (two-layer models) in the near-wall region. Our focus is on two-layer models and enhanced wall functions. There are also models that do not need special treatment near walls, which we will study as well. Two different simulation studies are presented in papers A and B. The first is an investigation of the gyroscopic effects that are present when the swirling flow in a hydropower draft tube is forced through a bend. These effects cause an asymmetrical flow distribution after the bend, which is thought to induce losses. A new design method is considered that tries to minimize the losses. In paper B, a sensitivity analysis is performed on three input parameters in a complex CFD simulation, the Turbine-99 Draft Tube test case. A two-level factorial design is used to assess the influence of surface roughness, radial velocity profile and dissipation length scale (last two at the inlet) on some engineering quantities such as the pressure recovery. Factorial design is shown to be an efficient, quantitative tool to recognize the important input parameters. It can suggest what experiments that should be carried out in order to get sufficient input data for simulations.