Secondary flow fields in Francis turbines : mapping and analyzing dynamics in rotor-stator interaction and draft tube flow with novel methods

Abstract

Hydropower, and especially Francis turbines, for electricity production has a history of more than 100 years and has proved to be one of the most efficient ways of utilizing renewable energy for electricity production. Yet, there are several problems to be solved regarding producing and running cost effective, high efficient and durable turbines. Secondary flow fields are all unwanted flow patterns present in the turbine. The major fluctuating flow fields in Francis turbines are caused by rotor‐stator interaction when the runner vane passes the guide vane wake and the swirling flow in the draft tube at off‐design operation. Such flow fields have a negative effect in terms of causing losses, vibrations, noise or damage to the turbine structure. The flow through a Francis turbine, especially at off‐design operation is not optimal, and is characterized by a dynamic and fluctuating flow pattern. It is difficult, but important to understand the behavior of the dynamics to better predict the negative effects of the fluctuating flows, and also in order to minimize or remove the unwanted effects by e.g. geometry modifying or flow control.

This work aims to introduce new methods helping to obtain a deeper understanding on the dynamics present in wake flow and in rotor‐stator interaction. It is investigated whether vortex generators, VGs, can have a positive effect on the wake with respect to rotor‐stator interaction. Experimental TRPIV (Transient Particle Image Velocimetry) wake data recorded at 10 000 Samples/sec from a cylinder in a stream at 1‐6 m/s and hydrofoils in a stream at 9 m/s are studied. Both plain hydrofoils, and hydrofoils where Vortex Generators – VGs are mounted are used in the study. The Reynolds number is in the range of 1.2∙104 – 7.3∙105. The velocity fields from both the cylinder and the hydrofoils are used as inlet boundary condition in a 2D CFD‐case simulating rotor‐stator interaction. The characteristic frequencies of the system are the vortex shedding frequency and the rotor passing frequency. The cylinder case shows that leading edge stagnation point moves with the vortex shedding frequency. The shedding frequency is also found in leading edge pressure and lift fluctuations, but the rotor passing frequency is found more dominant. Only the rotor passing frequency and its multiples is observed when looking into the same fluctuations for the hydrofoil case. Vortex shedding is only observed at Angle of Attack ‐ AoA=8 degrees for the hydrofoil. In terms of rotor leading edge pressure reduction, only AoA=0 and AoA=2 showed equal or slightly improvements with VGs. A time averaged wake is also used as input in the CFD for AoA=4 and it shows large amplitude reduction when looking at the leading edge pressure and lift oscillations.

Proportional Orthogonal Decomposition – POD is introduced to reveal flow structures and their corresponding energy in the wake flow. POD shows that up to 65 % of the total energy is located in the two first POD modes for the cylinder wake. Up to 18 % of the total energy is located in the two first modes for the hydrofoil at AoA=8 degrees where vortex shedding is present. The vortex generators prove to break the characteristic shedding pattern and move the early mode energy to later modes and reduce the energy content in the two first modes from 18 % to 8 % at AoA=8. At other AoAs only slight change in the energy distribution was observed due to the vortex generators.

The second part of the thesis is focusing on measurements and investigation of draft tube flow dynamics. Pairwise radial dynamic measurements of the swirling draft tube flow is done at the 25 MW Svorka power plant in Surnadal operating at 48 % load, at 6 radial and 7 angular positions. The data is analyzed with traditional methods as well as with POD. The pressure fluctuations were found to peak at R5, at 83 % of the radius. The Rheingans frequency is found to be the only distinctive peak, and is very dominant in the pressure signal. The POD addressed 52 % of the total pressure variance to the azimuthal mode 1. As the POD will pin point structures of high energy it is desired to be able to use this information in active flow control applications. This is difficult as the measurements in the draft tube are performed manually and are time consuming. The resulting eigenfunctions from the POD measured at one operating point can later be coupled with a single wall pressure transmitter through Linear Stochastic Estimation – LSE in order to give POD results for other operating points. This work presents the first step of required measurements in one operational point in order to verify the use of POD in draft tube flow. Further development of POD for use in active flow control through LSE will be a continuation of this work.