International Journal of Fluid Machinery and Systems DOI: 10.5293/IJFMS.2011.4.2.243 Vol. 4, No. 2, April-June 2011 ISSN (Online): 1882-9554

Original Paper (Invited)

Flow simulation and efficiency hill chart prediction for a Propeller turbine

Thi Vu1, Marcel Koller2, Maxime Gauthier1, Claire Deschênes3

1Andritz Hydro Ltd. 6100 Transcanadienne, Pointe Claire, H9R 1B9, Canada, thi.vu@andritz.com,

maxime.gauthier@andritz.com 2Andritz Hydro AG

Hardstrasse 319, 8021 Zürich, Switzerland, marcel.koller@andritz.com 3Laval University, Laboratory of Hydraulic Machines (LAMH)

1065 Avenue de la Médecine, Québec, G1V 0A6, Canada, claire.deschenes@gmc.ulaval.ca

Abstract

In the present paper, we focus on the flow computation of a low head Propeller turbine at a wide range of design and off-design operating conditions. First, we will present the results on the efficiency hill chart prediction of the Propeller turbine and discuss the consequences of using non-homologous blade geometries for the CFD simulation. The flow characteristics of the entire turbine will be also investigated and compared with experimental data at different measurement planes. Two operating conditions are selected, the first one at the best efficiency point and the second one at part load condition. At the same time, for the same selected operating points, the numerical results for the entire turbine simulation will be compared with flow simulation with our standard stage calculation approach which includes only guide vane, runner and draft tube geometries.

Keywords: Propeller Turbine, CFD, Flow Simulation, Performance prediction, Non-homologous geometry, Draft Tube

1. Introduction Due to the growing demand for hydro-electric energy production, the requirements on low-head hydraulic turbines are changing. The

need for increased power output and annual energy production of modernized and new power plants often involve the extension of the operating region of the turbines towards both full load and part load conditions. In these off-design operating regions, the flow in the turbine is characterized most of the time by time-dependent hydraulic phenomena, which are difficult to be simulated accurately by steady state flow computation. In a low head water turbine, the draft tube has to convert a high amount of kinetic energy of the flow leaving the runner which leads to a high energy loss in comparison with others turbine components. The highly swirling and decelerating flow in the draft tube makes the flow simulation of this component very difficult. Therefore performing flow simulation and predicting the efficiency of a low head water turbine for the whole range of operating conditions is a challenging task.

Andritz Hydro participates in the Consortium on Hydraulic Machines at the ‘Laboratoire de Machines Hydrauliques (LAMH)’ of Laval University in Québec, Canada. This research consortium aims at the creation of a comprehensive database of flow measurements in low-head water turbines for a wide range of operating conditions. In the first research project of the consortium, CRD AxialT, the flow in a propeller turbine model has been investigated in detail by model measurements on the university test rig [1]. Figure 1 shows the AxialT model and various locations of flow measurement in different operating points using different measurement techniques. These state-of-the-art techniques for measuring the flow in a hydraulic turbine have been developed and applied by the university [2], [3], [4], [5].

For the project partners, the huge set of steady and unsteady flow measurements in a low-head turbine model is a very valuable database to increase their knowledge of the flow phenomena in this type of turbines and to validate and improve their numerical flow simulation tools.

The AxialT turbine has a semi-spiral casing with two intake channels, 24 stay vanes, 24 guide vanes and a 6-bladed Propeller runner. The draft tube has a short cone, an unsymmetrical elbow and one pier. Special attention has been paid to the blade geometries of the old runner model. All 6 blades of the model were individually measured. As described by Nicolle et al. [6], the blade shapes of the AxialT model runner slightly differ from each other. The influence of these small differences in runner blade geometry could have an impact on

Accepted for publication April 27 2011: Paper number O11007S Corresponding author: Thi Vu, Hydraulic Engineering R&D, Andritz Hydro, thi.vu@andritz.com

This manuscript was presented at the 25th IAHR Symposium on Hydraulic Machinery and Systems, September 20-24, 2010, Politehnica University of Timisoara, Romania .

243

the result of the numerical flow analysis. As the experimental results are obtained with a runner model with 6 different blade geometries, we should perform the CFD simulation with a set of meshes representing the runner with all six different blade geometries. A simple approach which allows us to take into account all the 6 different blade geometries is to average them all and create new average blade geometry.

In the present paper we focus on the steady state flow computation. First, we will present the results on the efficiency hill chart prediction of the entire AxialT turbine using the average blade geometry and a discussion on the consequence of using different blade geometries for the CFD simulation. Secondly, the flow characteristics of the entire turbine will be investigated and compared with experimental data at different measurement planes. Two operating conditions are selected, OP3 near the best efficiency point and OP1 at part load condition. At the same time, for the same selected operating points, OP1 and OP3, the numerical results for the entire turbine simulation will be compared with flow simulation with Andritz standard stage approach which includes only guide vane, runner and draft tube geometries.

Fig. 1 CRD AxialT propeller turbine model & locations of flow measurements

Fig. 2 Normalized efficiency hill chart of the AxialT propeller turbine model

Fig. 3 LAMH main test rig with the AxialT model

2. Problem setup 2.1 Research environment

AxialT, the first project undertook by the Consortium on Hydraulic Machines, is centred on flow measurements in a propeller turbine model at the LAMH and on CFD analysis performed by the partners. A series of detailed measurements have been performed on the AxialT model by LAMH, to document the flow dynamic behaviour over a wide range of operation conditions. Their selection was based on the AxialT efficiency hill chart measured on the LAMH main test rig (see Fig. 2), on cavitation tests and characterisation of the vortex rope across the efficiency chart. Nine operating points were initially chosen for investigation of the unsteady velocity and pressure fields in various sections of the AxialT model: draft tube, runner, rotor-stator interface and intake. Three different heads and seven different openings were targeted to cover partial load, best efficiency and overload conditions. The influence of flow rate variation and guide vanes opening variation on the local flow conditions could thus be assessed. A tenth operating point was used for investigation of the cavitating vortex rope at partial load.

Test facility. The LAMH main test rig (see Fig. 3) consists of a classical closed-loop hydraulic facility with flow rate up to 1 m3/s, head up to 50 m, rotational speed up to 2000 rpm and a net power at the test section up to 170 kW. Its capacities permit incorporating regular model sizes of all types of reaction turbines, with either vertical or horizontal axis (Francis, Kaplan, propeller and bulb). A particularity of the loop resides in the use of an eddy current brake as power dissipater that can be mounted either horizontally or vertically. It is fully equipped with the best measurement systems available for testing the performances of models of hydraulic turbines according to the IEC 60193 standard [8] and for acquiring internal flow and pressure fields.

Advanced experimental investigations. Beside the standard instrumentation for model testing, LAMH uses laser-based optical methods to measure the unsteady velocity fields and a custom-made probe to measure the static and dynamic pressure fields in different sections of the turbine. Laser-doppler velocimetry (LDV) allows simultaneous point-wise measurements of two velocity components in a flow. The system, Dantec Dynamics, is composed of a 5.8W Argon-Krypton continuous laser source with output wavelengths of 488 and 514.5 nm, two-component laser beam delivery probes with integrated photo detectors, and data processors. Modular optical components (2 probes, beam expander, 5 front lenses) allow investigations at focal lengths ranging from 60 to 1000 mm, and provide increased flexibility in the design of experimental setups. An example of 2D-LDV probing downstream the runner hub on the AxialT model is given in Fig.4a.

A stereoscopic particle image velocimetry system (3D-PIV), also a Dantec Dynamics instrument, provides the three components of the velocity field for a given investigation area. This PIV system is composed of a 120mJ pulsed Nd:YAG laser source yielding 532nm wavelength beams, adapted light-sheet forming optics, and two CCD sensors cameras of 1280 x 1024 pixels operating in double-frame exposure mode. Figure 4b shows the experimental setup for 3D-PIV measurements in the conical diffuser of the AxialT turbine.

LAMH has extensive experience with the design and optimisation of high-quality optical interfaces, which is the critical parameter affecting the accuracy of LDV and PIV measurements given the complex geometry of hydraulic turbomachinery. Moreover, a range of seeding particles of various materials, geometrical characteristics and emission spectra are available for the laser-based instrumentation, giving flexibility in defining the experimental setups.

244

With respect to pressure field investigations, a custom-made unsteady pressure probe provides the total and static pressures, as well as the three components of velocity. Developed in house, the head of this instrument has an equivalent diameter of 8mm, and incorporates five miniature pressure transducers (Unisensor, 2 mm diameter, frequency up to 50 kHz), flush-mounted on its five faces [5]. Specific data processing procedures are applied to the raw unsteady data, and based on parametric calibration charts, to determine the flow direction, static and dynamic pressure components. Based on the calibration, the accuracy of the probe in terms of velocity is ±2% for angles smaller than 10°, and ±3.8% for an angular range of [–25°, +25°]. In terms of pressure, the accuracy is ±3.4% and ±5.8%, respectively, for the same angular ranges. Miniature sensors are also used for measuring the dynamic loading on the runner blades, through embedded pressure sensors linked to a telemetric data transmission system with 32 channels.

4a) LDV system at runner outlet 4b) PIV at runner outlet 4c) unsteady pressure probe at rotor stator interface

Fig. 4 Installation of instruments in the model

Velocity field investigations. For the purpose of this paper, the axial and tangential components of the velocity profiles will be presented, issued from 2D-LDV measurements at the guide vanes-runner interface, in 3 cross-sections downstream the runner, and in a cross-section at the draft tube outlet (see Fig. 5).

Figure 5 illustrates the velocity field synchronous with the runner at the rotor-stator interface and in the conical diffuser; the axial velocity is displayed in both sections, according to the Vz scale. Phase averaging was applied to the raw LDV data with respect to the runner rotation frequency. The simultaneous recording of the runner position, via an optical encoder, provided the time reference for the reconstruction of the periodic signal. Spectral analysis at the runner inlet and outlet also revealed the blade passing frequency and higher harmonics of the runner rotation in both sections. The circumferential discretization of 0.5 degrees on the averaged fluctuating velocity components allows assessing the individual influence of the blade passages.

Fig. 5 Axial velocity profiles measured by 2D-LDV at the best efficiency operation conditions: phase-average values at the rotor-stator interface and downstream the runner; average at the draft tube outlet

The spatial resolution in the radial direction is 5mm at the rotor-stator interface and 10mm downstream the runner, while the positioning error is maintained below 1% of the local cross-section diameter. A total of 60 000 acquisitions per position were performed for an estimated uncertainty on fluctuating velocities varying from 1% to 7% at the runner outlet, depending on the radial position. Zones close to the runner axis and near the wall exhibit steeper gradients and higher turbulence intensities, thus forcing higher uncertainties. These values are greatly reduced at the interface, where shorter integral time scales are encountered, and higher data acquisition rates have been observed with an uncertainty of 0,4% .

At the draft tube outlet, the spatial resolution is 20mm in the horizontal direction and 40mm in the vertical direction, with a maximum positioning error of 0.2% of the channel width. The mean axial velocity field for this section is displayed in Fig. 5, according to the Vx scale. The systematic errors, intrinsic to the LDV system, are 1.3%. The statistical convergence of the average values is attained for all positions, resulting in uncertainties below 2.5 % for the mean velocity, which accounts for velocity bias, angular bias and repeatability.

245

2.2 AxialT runner blade geometry

The geometry of the propeller model runner was measured by IREQ-Hydro Quebec. For more detailed information, please see [6]. Using our in-house runner blade geometry design tool, we have created new blade geometry by averaging the 6 individual blade geometries. Figure 6 shows the geometry deviation of each individual blade compared to the main averaged blade. The scale varies from -0.5% to +0.5% of the throat diameter. The average deviation of all blades is about 0.3%. The most deviated one is the blade #1 with a maximum value of -0.5% at the blade leading edge. For the model throat diameter of 380mm, this deviation corresponds to -1.9mm. The blades #3 and #4 have the least deviation of about 0.15 % which takes place at the leading and trailing edge regions. According to IEC code which allows a maximum of ±0.1% deviation [8], we could not use any geometry among the six blades to simulate accurately the AxialT turbine flow behavior. Beside the variation on the blade geometry, the six blades have different tip clearances with the shroud. The tip clearance varies from 0.03% to 0.12% of the throat diameter. We keep an average blade tip clearance of 0.07% for all flow simulations in this paper.

Fig. 6 Contour of geometry deviation of individual blade compared to the main average blade geometry

2.3 CFD setting for coupled steady-state simulations of entire AxialT propeller turbine model

The computational flow domain for CFD simulation in the entire AxialT turbine model, as shown in Fig. 7, comprises the semi-spiral casing, the stay vanes, one guide vane, one runner blade and the draft tube. Grid generation for the spiral casing and stay vanes was made with the commercial grid generator ANSYS ICEM-CFD providing tetrahedral elements with prism layers resolving the boundary layer near the walls. For other components of the turbine, guide vane, runner and draft tube, the grid generation was made with in-house automatic mesh generators providing H- and O-type hexahedral meshes. The guide vane is over-hanging from 20 degree opening to the maximum opening at full load. The gap configurations due to over-hanging guide vane and the runner tip clearance are taken care of by the mesh generator. Only one guide vane and one runner blade channel are generated for the computation. The complete computational grid of the entire propeller turbine simulations contains about 4×106

nodes. The generated meshes are intended to be used with k-epsilon turbulence model which requires a y+ value varying from 30 to 100 for the first node near solid wall. The CFD simulation for efficiency hill chart prediction uses our standard stage approach which includes only guide vane, runner and draft tube geometries (Fig. 8). In such case, the inlet region of the guide vane channel is not at the usual stay vane – guide vane interface, but it is placed further upstream allowing a uniform incoming flow from the inlet to fully develop. For the sake of simplicity, we call this standard set-up Stage2 because there are 2 stage interfaces, Guide vane – Runner and Runner – Draft tube, used in this computation.

The commercial flow solver ANSYS CFX v12.1 is used for performing the flow analysis. Steady-state time-discretization with a constant pseudo-time step and the so called ‘high-resolution’ space-discretization (mostly 2nd-order-accurate) has been applied. Turbulence is modeled by the standard k-ε model. The connections between different sub-domains – from casing to guide vanes, from guide vane to runner and from runner to draft tube – have been modeled by circumferential-averaging stage-interfaces. Two operating points at rated net head have been analyzed with these entire turbine simulations: OP1 in part load and OP3 near best-efficiency point, see Fig. 2. The flow rate measured in the model test has been specified at the inlet normal to the boundary surface. Averaged static pressure has been set at the outlet boundary at the end of the draft tube extension box. In this setting, the head, the hydraulic power at the rotating runner and so the calculated hydraulic efficiency result from the simulation.

246

3. Efficiency hill chart prediction

Ideally, the computational flow domain for CFD simulation of a hydraulic turbine should comprise the entire turbine assembly. For steady state simulation, in order to save CPU time, it is preferable that the computational flow domain would be divided into two flow domains. The first one would include casing and stay vanes. The second domain would include guide vane, runner and draft tube. There is an advantage of dividing the entire turbine into two computational flow domains. The casing and stay vanes assembly, which are fixed components, requires only one simulation for a given operating condition to determine the component head loss. For subsequent operating conditions, the head loss of the casing and stay vanes assembly can be calculated simply by assuming that their losses are proportional to the square of the flow rate. For the guide vane, runner and draft tube assembly, the flow analysis has to be performed for all operating conditions of interest with corresponding guide vane opening positions. This approach has been successfully validated for Francis runners and can be found in [7].

Fig. 7 Computational flow domain for full turbine simulation

Fig. 8 Computational flow domain for Stage 2 simulation

Fig. 9 Measurement plane locations near the runner

We have performed several CFD simulations to compute the efficiency hill chart of the AxialT turbine with different variations of the runner blade geometry. For most of the time, the computation was made for a range of guide vane opening from 20 to 44 degrees with an increment of 2 or 3 degrees. The n11-value used in the experimental investigation is 124. At the guide vane inlet, we specify a uniform flow and an inlet flow angle matching the flow orientation at the stay vane exit. The computation starts with a guessed flow rate. There is a procedure to iterate the flow rate for a specific guide vane opening during the course of the simulation until the computed head of the entire turbine matches with the prescribed turbine head. The turbine head is calculated by adding the useful work produced by the runner, and the head losses of all individual turbine components. The average blade geometry runner was chosen for the first calculation. At the beginning, the computation was performed for a turbine head of 10m which is our standard turbine head used for low head turbines. Then we performed a second computation for 7m turbine head which is the turbine head during the test in the laboratory. We obtained the same results from the computations with both turbine heads.

Fig. 10 AxialT turbine efficiency with different blade runner geometries

Fig. 11 Head losses of individual components with different blade runner geometries

Figure.10 shows the comparison of the predicted turbine efficiency against the experimental data for a wide range of operating conditions. The normalized flow rate varies from 0.75 to 1.15. The numerical prediction matches very well with the lab data in

247

terms of the efficiency level and the position of the best efficiency point. This good correlation validates our approach of using the average blade geometry to represent a runner with 6 different blades. Figure 10 shows also the efficiency prediction for blade #1 which has the largest geometry deviation in the group. It is interesting to note that the position of the BEP of blade #1 is shifted to the left about 4% compared to the one of the average blade and the efficiency at the BEP is slightly higher of about 0.25 % compared to the average blade efficiency. For blade #4 which has the least geometry deviation in the group, the shift of the BEP to the left is smaller, about 2%. For the last computation, we have chosen blade #2 which has a positive geometric deviation at the blade trailing edge region as opposed to the two blades #1 and #4, which both have a negative deviation at the blade trailing edge. This time the BEP location of blade #2 is shifted to the right compared to the one of the average blade.

The head loss of the entire turbine can be broken into head losses of individual components as shown in Fig. 11 for the whole range of the turbine operating conditions. It indicates clearly that the shape of the efficiency hill chart of a low head turbine is governed by the performance curve of the draft tube. While the runner loss varies smoothly over the wide range of operating condition, the head loss of the draft tube varies sharply near the BEP. We can find that the position of the lowest energy loss in the draft tube corresponds with the location of the BEP of the turbine as shown in Fig. 10. The location of the lowest draft tube loss associated with the Blade #1 runner is also shifted 4% to the left compared to the lowest draft tube loss associated with the average blade runner. Figures 12 show the axial and tangential velocity profiles at the draft tube inlet for 3 different operating conditions; at the best efficiency point and at the two computed extreme off-design conditions. At the BEP, the tangential velocity profile is similar to a solid body swirl profile (Fig. 12b). At full load condition (Fig. 12c), 44 degree guide vane opening, there is little co­rotational swirl near the periphery and large contra-rotating flow near the draft tube center. Figure 12a shows the axial and tangential velocity profiles for a very part load condition, 17 degree guide vane opening. As observed, the tangential velocity component has same amplitude as the axial component at the periphery and surpasses it toward the draft tube center. This explains a high rise of draft tube loss for this condition, 4 times higher than the draft tube loss at BEP.

The head loss of the blade #1 runner is overall slightly smaller then the one of the average blade runner. This explains the higher efficiency of the blade #1 runner. The loss of the casing-stay vane assembly, which is similar for all blade runner geometries, increases with the flow rate Q11. On the contrary, the guide vane head loss decreases with flow rate Q11 due to a small flow passage between the guide vanes at part load condition. The guide vane loss is quite similar for all blade runner geometries and it is shown here only for the average blade computation.

12a) very part load (α = 17°) 12b) BEP (α = 33°) 12c) full load (α = 44°) Fig. 12 Axial and tangential velocity profiles at 3 different operating conditions

4. CFD simulations of the entire AxialT propeller turbine model The CFD simulation in the entire turbine model geometry including casing, stay vanes, guide vanes, runner and draft tube is

performed for two selected operating points, OP1 at part load condition and OP3 near the best efficiency point. The OP1 condition corresponds to the guide vane opening of 25 degrees and the OP3 condition corresponds to the guide vane opening at 33 degrees. For this computation, we imposed the turbine flow rate obtained from the model test with a uniform velocity distribution at the casing inlet. The turbulence intensity was set to 2% at the inlet. Concurrently with the full turbine computation, we performed CFD simulations with the Stage2 computation domain as described above for the same operating conditions, OP3 and OP1. We used the same flow rate imposed by the model test with a uniform flow angle of 45 degrees at the inlet of the guide vane. The turbulence intensity was set to 1% at the guide vane inlet.

The following is the comparison of the CFD results obtained from both setups against the experimental data. Figure 9 shows the location of several planes upstream and downstream of the runner used for comparison: the STV-GV interface (r = 0.25m), the GV-RN Interface (corresponding to the measurement plane #3) and the runner outlet/draft tube cone planes (Planes #5a, #5b and #5c). In all figures, the velocity has been normalized by the average axial velocity at the turbine throat.

4.1 Results and discussion for operating point OP1 at part load (α = 25°) At the STV-GV interface (Fig. 13), the velocity profiles from the two flow simulations are plotted in order to verify if the

imposed uniform flow at the inlet for the Stage2 calculation is valid. A good correlation is obtained for the radial component distribution suggesting that the position of guide vane inlet of the Stage2 simulation is adequate.

248

Fig. 13 Axial and tangential velocity profiles at the STV- Fig. 14 Axial and tangential velocity profiles at the GV-RN GV interface – OP1 interface Plane 3 – OP1

Fig. 15 Axial and tangential velocity profiles at the DT inlet Fig. 16 Axial and tangential velocity profiles under the hub Plane 5a – OP1 Plane 5b – OP1

Fig. 17 Axial and tangential velocity profiles inside the DT Fig. 18 Turbulence intensity profiles inside the DT cone cone Plane 5c – OP1 Plane 5c – OP1

249

Overall, for the OP1 condition, the results from the full turbine and Stage2 simulations match quite well with the phase-averaged velocity measurements. At the GV-RN interface (Fig. 14), the full turbine simulation correctly predicts the phase-averaged axial velocity profile, while slightly under-predicting the tangential velocity of the flow. The Stage2 simulation predicts the velocity profiles quite well, although it tends to overshoot the tangential velocity and to under-predict the axial velocity near the hub, with the reverse phenomenon at the shroud.

The velocity profiles at the 5a, 5b and 5c measurement planes (Fig. 15, 16 and 17) show that neither type of simulation can be said to be better predicting the flow in the draft tube cone. At 5a (Fig. 15), the CFD velocity profiles more or less match up to the measured data. However, downstream planes 5b and 5c show that the predicted tangential velocities are lower than measured. Also, while the axial component profiles match up pretty well over most of the 5b and 5c planes, the predicted behavior under the hub did not match up very well to measured data, even to the point where the CFD results show a large region of counter-flow under the hub, at the 5c plane. Figure 18 shows the distribution of the turbulence intensity at the plane 5c. The numerical results from both flow simulations are about 40% of the experimental data.

19.a Center plane - Full turbine simulation 19.b Center plane - Stage2 simulation

19.c Mid-Elbow plane - Full turbine simulation 19.d Mid-Elbow plane - Stage2 simulation

19.e End Elbow plane - Full turbine simulation 19.f End Elbow plane - Stage2 simulation

Fig.19 Comparison of computed velocity contour and velocity vectors at different draft tube section planes – OP1

250

Fig. 19 shows a qualitative comparison of the velocity contour and velocity vectors obtained with the full turbine and Stage2 simulation. Both simulations yield a very similar flow behavior at different draft tube section planes. For this part load flow condition with high swirl intensity, a large flow recirculation zone is observed at the draft tube center plane. The strong swirling flow is transported to the draft tube elbow as shown in 19.c and 19.d for secondary flow in the mid-elbow plane. At the end of the elbow where the draft tube flow enters into two pier channels, the swirling flow still remains at the left channel entrance. Finally, the velocity contours at the draft tube outlet obtained with both simulations are very similar and compare well with the experimental velocity contour (Fig. 20 and 21). The flow pattern is very similar as the one observed at the draft tube end-elbow section plane. The measured mass flow distribution for the two draft tube channels is 23.1% and 79%. We obtain a distribution of about 22% - 78% for the full turbine and 21.5% - 78.5 % for the Stage2 simulation.

Full turbine simulation Stage2 simulation

Fig. 20 Experimental velocity contour at Fig. 21 Computed velocity contour at Draft Tube Outlet – OP1 Draft Tube Outlet – OP1

4.2 Results and discussion for operating point OP3 near BEP (α = 33°)

At the STV-GV interface for the operating point OP3 (Fig. 22), we find the same similarity as observed for the operating point OP1. Figure 23 shows the predicted axial and tangential velocity profiles at the GV-RN interface. While both simulations are very close to the measured axial velocity profile, neither simulation predicts with precision the measured tangential velocity profile. The full turbine solution under-predicts while the Stage2 over-predicts. The velocity profiles at the 5a plane (Fig. 24) show that the full turbine simulation is a better predictor of both the axial and tangential measured velocity profiles than the Stage2 simulation. The Stage2 solution under predicts the tangential velocity profile near the draft tube wall.

The same tangential velocity defect is observed for the planes 5b and 5c as shown in Fig. 25 and 26. At the same planes 5b and 5c, the full turbine velocity profiles match up well with the measured velocity profiles near the draft tube wall while it is rather the Stage2 simulation that seems to be better at predicting the velocity profiles near the hub region. One noteworthy difference between the two simulations is the inability for the full turbine simulation to predict the “surge” in tangential velocity near the center (at about r=0.02-0.03 m), which the Stage2 simulation has no trouble catching. Also, the full turbine simulation predicts a large area of flow recirculation directly under the hub (Fig. 25), which we found surprising because of the measurement plane’s proximity to the hub. Figure 27 shows the distribution of the turbulence intensity at the plane 5c for the OP3 condition near the BEP. It is surprising to see that the turbulence intensity profile obtained from the full turbine computation is several times smaller compared to the experimental data and the result from Stage2 simulation. This could explain the different results from the two simulations at the hub region

Fig. 28 shows a qualitative comparison of the velocity contour and velocity vectors obtained with the full turbine and Stage2 simulation. For the BEP flow condition, there is no flow recirculation in the draft tube cone. Only the Stage2 simulation predicts a more important wake under the runner hub (Fig. 28a & 28b). Also at the mid-elbow section plane, the intensity of the secondary flow is much less important than the one of part load condition. At the end-elbow section plane, the difference in flow rate for the two pier channels is not very important with larger proportion for the right channel. Figures 29 and 30 show a comparison of the velocity contours at the draft tube outlet. The measured mass flow distribution for the two draft tube channels is 38.1% and 52.8%. The values are the percentage of mass flow in one channel (measured and integrated from LDV data) compared to total mass flow measured on test rig. Since LDV measurements did not cover the full cross-section, the sum of both values is not 100%. While the two types of simulation are moderately close in terms of mass flow distribution, about 36% - 64% for the full turbine and 32% ­68 % for the Stage2, the flow pattern show little similitude between the 2 flow simulations and the experimental data.

251

Fig. 22 Axial and tangential velocity profiles at the STV-GV Fig. 23 Axial and tangential velocity profiles at the GV-RN interface – OP3 interface Plane 3 – OP3

Fig. 24 Axial and tangential velocity profiles at the DT Inlet Fig. 25 Axial and tangential velocity profiles – Plane 5b – Plane 5a – OP3 OP3

Fig. 26 Axial and tangential velocity profiles Fig. 27 Turbulence intensity profiles inside the DT Cone inside the DT cone (Plane 5c) – OP3 Plane 5c – OP3

252

28.a Center plane - Full turbine simulation 28.b Center plane - Stage2 simulation

28.c Mid-Elbow plane - Full turbine simulation 28.d Mid-Elbow plane - Stage2 simulation

28.e End-Elbow plane - Full turbine simulation 28.f End Elbow plane - Stage2 simulation

Fig.28 Comparison of computed velocity contour and velocity vectors at different draft tube section planes - OP3

Full turbine simulation Stage2 simulation

Fig. 29 Experimental velocity contour at Fig. 30 Computed velocity contour at Draft Tube Outlet – OP3 Draft Tube Outlet – OP3

5. Conclusion In the present paper, we have presented flow simulations of a low head Propeller turbine at various design and off-design

conditions. We have demonstrated that creating an average blade runner to represent a model runner with different geometry variation is a valid and simple approach allowing us to predict correctly the efficiency hill chart of this particular runner and we

253

have shown the consequence of geometry deviation on the efficiency hill chart prediction. It is crucial to model the correct geometry, even small deviations in runner blade geometry could lead to inaccurate results (e.g. when taking blade 1 instead of the averaged blade). Obviously, we will perform comparative CFD simulations with a computational domain including 6 different blade geometries for further validation. Also, we have performed comparative simulations with the full turbine and Stage2 flow domains. Both computations give relatively similar results, but the difference found in the draft tube flow prediction in OP3 has to be investigated. One of the possible reasons is the difference in the turbulence intensity level developed at the runner outlet which leads to different results in the velocity profile below the hub.

The steady state CFD analysis shows reliable results for the analysis of global turbine characteristics for a range of -25% to +15% of the flow rate from the BEP, as demonstrated in our efficiency hill chart prediction, given that the appropriate geometry is used. However, details of flow patterns (e.g. swirl and backflow in hub wake) are not exactly predicted. In such case, an unsteady simulation is necessary, especially in the draft tube, where time-dependent flow phenomena with different timescales exist, which has major impact on the performance of low head water turbines. However, results of these investigations will be presented in a subsequent publication.

Acknowledgments The authors would like to thank the participants on the Consortium on Hydraulic Machines for their support and contribution

to this research project: Alstom Power & Transport Canada Inc., Andritz Hydro, Edelca, Hydro-Quebec, Laval University, NRCan, Voith Hydro Inc. Our gratitude goes as well to the Canadian Natural Sciences and Engineering Research Council who provided funding for this research. A special thanks to IREQ – Hydro Quebec having provided us the AxialT geometry in a requested specific format. The in-house automatic mesh generators for distributor, runner and draft tube are from the project Gmath, acollaborative R&D project between École Polytechnique de Montréal and Andritz Hydro Ltd.

Nomenclature Aref Area of draft tube outlet section [m2] Q Flow rate [m3/s] BEP Best efficiency point Q11 Unit flow rate Q11 = Q/D2H0.5

cx Horizontal velocity component in direction of Ox [m/s] r Radius [m] cref Reference velocity at DT outlet cref = Q/Aref [m/s] RN Runner CS Spiral casing STV Stay vanes D Throat diameter of the turbine [m] vr Radial velocity component [m/s] DT Draft tube va Axial velocity component [m/s] GV Guide vanes vt Circumferential velocity component [m/s] n11 Unit speed n11 = nD/H0.5 w Axial velocity component [m/s] OP Operating point α Guide vane opening angle [°] Ox Horizontal reference axis pointing towards tail water η Hydraulic efficiency Oz Vertical reference axis, turbine axis ref Index referring to the operating point near best-

efficiency point

References [1] Deschênes C., Ciocan G. D., De Henau V., Flemming F., Huang J., Koller M., Arloza F. N., Page M., Qian R, Vu T. C., 2010, “General overview of the AxialT Project: a partnership for low head turbine developments,” 25th IAHR Symposium on Hydraulic Machinery and Systems, Timisoara, Romania. [2] Gagnon J. M., Iliescu M., Ciocan G. D., Deschênes C., 2008, “Experimental Investigation of Runner Outlet Flow in Axial Turbine with LDV and Stereoscopic PIV,” 24th IAHR Symposium on Hydraulic Machinery and Systems, Foz do Iguassu, Brazil. [3] Beaulieu S., Deschênes C., Iliescu M., Fraser R., 2009, “Flow Field Measurement Through the Runner of a Propeller Turbine Using Stereoscopic PIV,” 8th International Symposium on Particle Image Velocimetry – PIV09, Melbourne, Australia. [4] Gouin P., Deschênes C., Iliescu M., Ciocan G. D., 2009, “Experimental Investigation of Draft Tube Flow of an Axial Turbine by Laser Doppler Velocimetry,” 3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno, Czech Republic. [5] Duquesne P., Iliescu M., Fraser R., Deschênes C., Ciocan G. D., 2010, “Monitoring of velocity and pressure fields within an axial turbine,” 25th IAHR Symposium on Hydraulic Machinery and Systems, Timisoara, Romania. [6] Nicolle J., Labbé P, Gauthier G., Lussier M., 2010, “Impact of blade geometry differences from CFD performance analysis of existing turbines,” 25th IAHR Symposium on Hydraulic Machinery and Systems, Timisoara, Romania. [7] Thi C. Vu, Safia Retieb, 2002, “Accuracy assessment of current CFD tools to predict hydraulic turbine efficiency hill chart,” 21st IAHR Symposium on Hydraulic Machinery and Systems, Lausanne, Switzerland. [8] IEC Code 60193 - Hydraulic turbines, storage pumps and pump-turbines – Model acceptance tests, Second edition.

254