 |
Transient CFD Analysis of a Pelton Turbine
|
|||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 1 - Rovesca hydropower plant |
Franco Tosi Meccanica
Franco Tosi Meccanica S.p.A. (FTM) operates in the power generation industry since early 1900. The Italian company headquartered in Legnano (MI) is specialized in the design, manufacture, installation and start-up of medium to high power steam and hydraulic turbines ranging from 1 to 700 MW. Typical applications address power plants as well as civil and industrial cogeneration systems for both Italian and foreign markets. Furthermore, a wide range of services is performed on numerous FTM installations all over the world.
Pelton Turbines at a glance
In recent years, global warming and environmental issues have led to a continuous growth in demands for green energy supplies. Hydropower is certainly one of the major resources in green power generation. New markets, also in developing countries, demands linked to the repowering of plants, maintenance and reconditioning (especially in the Western markets), explain the growing efforts and investments in R&D for hydropower. The common goal is to provide more efficient and more reliable systems for the future.
Pelton turbines play a primary role in hydraulic machinery as they represent one of the most efficient types of water turbines. Invented by L.A. Pelton in the 1870s, they have been the subject of many studies focused on increasing performances in terms of mechanical efficiency and reliability.
Before the advent of computers, researchers mainly considered dynamic similarity laws and experimental tests although results where not always easy to obtain and to analyze. This was particularly true for the evaluation of the fluid dynamics of the turbine buckets due to their complex shape, 3D geometry and the multiphase and unsteady nature of the flow.
Nowadays, CAE, and more specifically CFD, offers researchers and engineers sound opportunities in the research and development for Pelton turbines. These state-of-the-art technologies allow engineers not only to evaluate the three dimensional fluid flow on the bucket, but also to quickly estimate the effects of modifications on the machine design by applying a what-if scenario approach.
Rovesca Hydropower Plant
The Rovesca power plant is a recently restructured hydropower plant that is situated near Verbania in the Italian Alps. It is illustrated in Figure 1 and Figure 2.
 |
Fig. 2 - The machine object of study |
In order to increase the plant’s efficiency and reliability, the management asked FTM for support in reconditioning the systems with the primary objective to renew the Pelton turbine.
The machine, a 2 injectors Pelton turbine whose rating is reported in Table 1, is based on a horizontal-axis and 18 buckets.
The Mathematical Model
A hydroelectric power plant represents a complex system. From a fluid-dynamic point of view it is not strictly confined to the generating station, but also comprises the penstocks, the tailrace and all the necessary devices required to control the flow, such as, for example, the control valves.
Even if the creation of a model that includes all these entities would be possible, such an approach would be inopportune,both in terms of domain size and model complexity.
In the present case, thanks to prior investigation carried out by FTM on the penstock ducts flow, the computational domain at the injectors outlet sections could be reduced. The obtained domain geometry is shown in Figure 3.
We should note that the injectors are just modeled as voids in the fluid domain: the water inlets correspond to the jets’ vena contracta cross-sections whose details are known thanks to the measurement.
With the axial midplane symmetry of the machine, the model could be conveniently halved (Figure 4). No further computational domain reductions are possible due to the injectors angular displacement (80 deg).
As a next step, the resulting domain has been decomposed into the statoric domain (Pelton wheel housing) and the rotor domain (Pelton wheel). Each domain has been discretized independently using ANSYS ICEM CFD.
The stator frame discretization led to 1M cells, a full hexa mesh structure as shown in Figure 5.
 |
 |
 |
| Fig. 3 - Turbine geometry with simplified statoric frame | Fig. 4 - Half turbine model | Fig. 5 - Statoric frame mesh |
The rotor frame has been further reduced to a single bucket. Two different discretization approaches have been deployed: tetrahedral meshing (Octree algorithm with prismatic boundary layer extrusion) and structured hexahedral meshing (Figure 6 and Figure 7) which allowed mesh sensitivity studies.
During the entire discretization process, great care has been taken in treating the air/water free surface as well as the interfaces between the domains and the bucket boundary layers in order to ensure a proper resolution of the momentum transfer processes occurring in these regions. After all, they have a great impact on the wheel torque and, ultimately, on the efficiency of the machine.
Details of the obtained mesh are reported in Table 2.
The model setup in terms of boundary conditions is schematically illustrated in Figure 8. It consists of two fluid inlets, one for each injector, an opening outlet, to allow possible fluid recirculation occurring at the discharge section, a symmetry plane and, finally, two mesh interfaces in order to ensure fluid-dynamic continuity between stator and rotor frames.
 |
Fig. 6 - Rotoric sinle bicket mesh. Structural mesh (left), and unstructured tetra-prism mesh (right) |
The flow is multiphase consisting of liquid water and air. We assumed that both are incompressible and unmixable and therefore, a multiphase model has been adopted. With the purposes of the analysis in mind, a homogeneous mathematical treatment for both hydrodynamic and turbulence quantities has been chosen.
Furthermore, the adoption of an interface compression algorithm allowed a better resolution of the air/water interface.
The flow is also assumed to be fully turbulent, hence a suitable turbulence model is needed. In the present study, two different RANS models have been employed: k-ε and SST k-ω with an automatic boundary layer treatment based on the logarithmic wall law.
 |
 |
 |
| Fig. 7 - Fully hexahedral rotoric frame assembly | Fig. 8 - Boundary conditions | Fig. 9 - Correlation between average mechanical efficiency and numerical discretization scheme |
Steady state as well as transient analyses have been carried out. The former have been performed to initialize the flow, while the latter confirmed the actual machine performances including unsteady phenomena, and completely resolved all fluid jets-buckets interactions.
All the final transient runs adopted High Resolution schemes with second order spatial accuracy ensuring optimal gradients resolution.
Solution
The simulations have been remotely run on the main EnginSoft HPC linux cluster. The cluster is powered by 324 cores based on a 64bit Intel Xeon cpu architecture and adopts the CentOS operating system. as well as Perceus, an enterprise management suite for HPC clusters deployment and management. Jobs submitted by users are handled by a Sun Grid Engine that manages queues in order to reach and maintain the highest possible availability and usability of the cluster’s computational resources.
A powerful parallel data storage infrastructure, capable of continuous data flow of 400kB/s, completes the system. With more than 15 TB of total capacity it allows users to perform even computations with high output needs.
Based on these great computational resources, it has been possible to perform multiple parallel runs testing different set-ups. The final runs employed between 16 and 40 cores, allowing an impressive wall clock time speed-up of our computations.
Results
The overall machine performances in terms of mechanical efficiency are reported in Figure 9 where it is also possible to correlate the efficiency trend respect to the adopted discretization scheme: analyses are performed with increased complexity and accuracy in terms of gradients resolution so that each run presents higher average efficiency than the previous one. Highest performances are reached by the fully hexahedral mesh.
 |
 |
Fig. 10 - Buckets contributes to total hydraulic torque (black time) |
Fig. 11 - Correlation between single buckets' torque contributes and their fluid-dynamic state |
Figure 10 shows the contribution of each single bucket to the total hydraulic torque (black line). Such contributions are of great interest to FTM’s engineers as they can link them to the fluid dynamic state of the buckets (Figure 11). In this way, they can determine when the momentum transfer is not optimal and where to apply possible improvements to the bucket’s shape.
The water path on the buckets is reported in Figure 15: thanks to the tools available in ANSYS CFD-Post, it was possible to evaluate the water layer thickness and distribution. Such details are important for Pelton designers, as they are extremely difficult to obtain experimentally but quite easy to retrieve with CFD. Another example of the benefits of CFD is presented in Figure 12, where we can verify the absence of interactions between the two water jets.
 |
 |
Fig. 12 - Distinct tracers in each water jet allow to verify the absence of interaction between them |
Fig. 13 - Water volume fraction on the turbine's axial midplane |
Maps and contours of different quantities on the machine’s midplane are shown in Figure 13 – Figure 16. In particular, Figure 16 helps to estimate water drops recirculation in the near wheel region.
The analyses also took the cavitation phenomena into account: Figure 17 highlights a bucket surface area with high cavitation tendency.
 |
 |
 |
Fig. 14 - Water absolute velocity on the turbine's axial midplane |
Fig. 15 - Water volume fraction on a bucket frontal cutplane | Fig. 16 - Water drops recirculation displayed on the turbine's axial midplane |
Finally, monophase analyses were performed with air only returned from the aerodynamic rotor losses.
Conclusions
For the renovation and upgrade of a hydropower plant, a fluid-dynamic analysis of a full-scale model of a Pelton turbine has been carried out.
The use of CFD has enabled the engineers to overcome the machine’s intrinsic complexity and to achieve the following:
- Various turbine performances could be analyzed, for example: hydrodynamic torque and power output as well as the magnitude of these quantities’ and their fluctuations.
- Detailed resolution of the water jets interaction with the wheel’s buckets.
- Successful verification of the absence of mutual jet interaction.
- Detailed 3D study of the water path on the bucket and of the fluid layer thickness.
- Linkage of each single bucket’s fluid-dynamic state with its actual contribution to the total shaft torque.
- Verification of the entity of the stator frame water recirculation phenomena in the wheel proximity
- Evaluation of the cavitation tendency of the machine
- Estimate of idle losses with free-to-air running wheel.
 |
Fig. 17 - Detail of a bucket surface area with high cavitation tendency |
Franco Tosi Meccanica has benefited from the work performed and the applied technologies as indicated by the described results. Moreover, the company could model strategic technological developments. HPC resources played a crucial role too, speeding-up computing time thanks to EnginSoft’s resources.
Future project work will focus on evaluating more complex non-homogeneous multiphase models, post-processing macros to easily calculate flow angles on buckets and, finally, new partitioning algorithms, such as the Coupled Domain Multipass Partitioning, available in ANSYS CFX 12 and applied to increase calculation speed and robustness.
References
ANSYS CFX-Solver Theory Guide
ANSYS CFX-Solver Modeling Guide
G.Minelli (1980), Macchine idrauliche, Ed. Pitagora
Perrig, A. (2007). Hydrodynamics of the Free Surface Flow in Pelton Turbine Buckets, PhD Thesis, EPFL
Zoppé, B., Pellone, C., Maitre, T., Leroy, P. (2006), Flow Analysis Inside a Pelton Turbine Bucket, ASME
Luca Brugali, Lorenzo Bucchieri, Cristian Catellani
EnginSoft S.p.A.
Emanuel Pesatori, Giorgio Turozzi
Franco Tosi Meccanica S.p.A.
www.francotosimeccanica.it
Article published in the Magazine: EnginSoft Newsletter Year 7 n.1