1 3D hot-wire measurements of a wind turbine wake Pål Egil Eriksen PhD candidate, NTNU/NOWITECH Per-Åge Krogstad NTNU
2 Outline of the presentation Experimental setup Measurement technique Time averaged results Phase-locked-averaged(PLA) results Possibilities for further analysis of the data Conclusions
3 Experimental setup (1/2) Exact same setup which was used in Blind Test 1 [1] Turbine positioned 4D from the entrance of the test section Test section 11.2m x 1.8 m x 2.7 m Allows for measurements 5D downstream of the turbine Data collected at 1D,3D & 5D for λ R = 6 along a horisontal line. Equipped with a balance and a traverse system Turbulence level 0.3 % Figure 1: Upstream view of the windtunnel [1] Blind test calculations of the performance and wake development for a model wind turbine. Krogstad and Eriksen, Renewable Energy, 2013
4 Experimental setup (2/2) Wind turbine model Diameter: 0.9 m Hub height: 0.8 m Re tip: ~100000 at λ R =6. Peak efficiency ~45% at λ R =6. Operated at a constant rpm using a frequency converter. Instrumentation: Torque sensor, rpm measurement using photo cell & slip rings. Photo cell and constant rotational speed makes phase locked averaging possible. Figure 2: Model turbine Figure 3: Blade profile (NREL S826)
5 Measurement technique (1/2) CTA hot wire anemometry 2.5 μm wire -> capable of high frequency response Figure 4: Sketch of crosswire Blind test 1 Used a single crosswire probe Consists of two wires Resolves two velocity components simultaneously Neglects cooling velocities normal to the plane of interest Can not resolve all shear stresses and third order moments Figure 5: Crosswire mounted on traverse in wind tunnel
6 Measurement technique (2/2) Current experiment Probe(hereafter called 2xw-probe) consisting of two cross wire probes measuring in orthogonal planes. Resolves all three components of the velocity vector Solved using an iterational procedure where binormal cooling is taken into account Probe crossection ~ 2mm Resolves all turbulent stresses Figure 6: Sketch of 2xw-probe
7 Time averaged results (1/2) Velocity defect Quite good match Deviation in the freestream of the order of 2-3% Probe rotation has a minor effect Figure 7: Velocity defect at x/d=1 for λ R =6
8 Time averaged results (2/2) Turbulent kinetic energy Quite good match Some deviation near the peak. Could be due to: Deviation in pitch angle Difference in probe response to flowfield Bump at z/r = -1.18. Why? Phase-locked average of the data can give us the answer. Figure 8: Turbulent kinetic energy at x/d=1 for λ R =6
9 Phase locked average (1/4) Averaging with respect to rotor position Position is determined using the rotational speed and the photo cell PLA of turbulent kinetic energy Reveals position of tip vortices. Shows that the tipvortex of one blade is located at a different radial coordinate. May explains the bump in Figure 8. Figure 9: PLA of turbulent kinetic energy at x/d=1 for λ R =6
10 Phase locked average (2/4) Can investigate how the presence of the vortices affects the mean velocity field Figure 10: Vector plot of [U_radial,U_tangential] at x/d=1 for λ R =6. Overlapped with axial velocity contours.
11 Phase locked average (3/4) The turbulence level in the tip vortex region is dominating in the wake(as shown in figure 9) PLA can also be used to reveal more of the internal structure of the wake. By plotting the axial normal stress on a logarithmic scale the turbulence produced by the boundary layer on the blade can also be visualized. Can also see a peak in the centre with increased turbulence intensity. Figure 11: PLA of the streamwise normal stress at x/d=1 for λ R =6. Logarithmic z-scale.
12 Phase locked average (4/4) A close up of the radial velocity reveals a 3p variation in the centre region Could also be seen in Figure 9 Figure 12: PLA of the radial velocity at x/d=1 for λ R =6. Figure 13: PLA of the radial velocity at x/d=1 for λ R =6. Centre region
13 Other possibilities The dissipation rate ε can be estimated, eg. from the dissipation spectrum. Relevant information for numerical modelers. Investigation of isotropy Triple correlations can yield information which can be useful for estimating terms in the transport equations for turbulent kinetic energy. Figure 14: Example of dissipation spectrum obtained at x/d = 1. Not normalized.
14 Conclusions The new results match quite well with the old blind test results. Phase locked average can reveal a lot of information about the structure of the wake, which it is not possible to find from time averaged measurements. There are many possibilities for further analysis on the dataset.