Redesign of an existing hydropower draft tube

Rehabilitation and modernisation of old constructions are important for a contemporary energy market. Among the renewable energy sources, hydropower has an eminent potential for further improvements since a great number of the hydropower plants are ageing and are as well often run at off-design conditions. An important part of a hydropower plant (low and medium headed) is the hydraulic turbine draft tube that contributes to a large portion of the hydraulic losses. The purpose of the draft tube, often being a curved diffuser connecting the runner to the outlet, is to recover kinetic energy and thus creating an artificial head. Traditionally the design has been based on model tests and simplified analytic methods. Today and in the future Computational Fluid Dynamics (CFD) in combination with computer optimization will be used more frequently as a design tool. The numerical prediction of the flow field in the draft tube is however challenging, caused by its complex flow features e.g. unsteadiness, swirl, separation etc. Therefore several numerical difficulties have to be solved before it can be applied routinely in product development. One of the key issues in this context is the turbulence modelling. Here the flow field is analyzed and validated with previously performed measurements on two draft tube geometries, a sharp-heel draft tube and a modification of it (where the sharp heel is smoothed). Both steady and unsteady simulations are performed, with the standard k-epsilon turbulence model as well as the SST turbulence model. The focus is set on the alteration in the pressure recovery factor and the overall flow field as a function of the shape of the draft tube and the implemented turbulence model. The steady and unsteady CFD simulations performed with the standard k- epsilon turbulence model yield about the same result. To exemplify, the difference in the pressure recovery factor between these simulations is much less than 0.001%. The main difference is that the unsteady simulation required less CPU-time as compared to the steady ones. The improvement in the pressure recovery between the original and the modified geometry is also small, about 0.006%. This can be compared to the experiments where the efficiency of the system improved with about 0.5%, indicating that the pressure recovery, as defined, should increase even more. The CFD simulations with the SST turbulence model did not converge in a proper manner although several attempts were made to improve the convergence. The reasons to this can be found in the applied inlet boundary conditions, the generated grids and/or in the location of the outlet boundary.Validerat; 20101217 (root