Community Research and Development Information Service - CORDIS

Final Report Summary - HYPERBOLE (HYdropower plants PERformance and flexiBle Operation towards Lean integration of new renewable Energies)

Executive Summary:
The 20-20-20 strategic energy policy decided by the European Union and its accompanying Renewable Energy Directive is leading towards a dramatic change in the means of production of energy, characterized by a massive penetration of alternative renewable energies such as wind and photovoltaic (PV) energy sources. With the ratification of the Paris Agreement on climate change (UNFCCC COP21), this movement has been reinforced, which is illustrated by the proposal published by the Commission in November 2016 to ensure a target of a least 27% renewable energies of the final EU energy consumption by 2030.
To achieve a lean integration of New Renewable Energy sources (NRE) within the power network, it is necessary to ensure that power stations have enough storage capacity as well as primary, secondary and tertiary grid control capabilities to be able to quickly establish energy parity between generation and consumption. Hydropower stations play an integral part in grid stability, as by design they are capable of immediate power generation for peak power surges and can provide ancillary services. Additionally, pumped storage power stations provide a power system to store large amounts of electricity, with a full cycle of pumping and generation that can achieve unrivalled efficiencies of at least 80%.
To this end, the HYPERBOLE consortium has accomplished important milestones and objectives for the enhancement of the capability of hydropower stations to function over larger operating ranges and with a faster response time.
One of the main milestones achieved by the HYPERBOLE project was the development of a complete methodology for predicting and assessing the dynamic behaviour of hydropower stations operating outside of their rated operating ranges and in transient conditions. In such conditions, hydropower units develop flow instabilities, putting at risk the stability of the hydro-mechanical system and, in worst-case scenarios, the electrical transmission system. This methodology requires proper one-dimensional (1-D) modelling of the hydroelectric system and the hydro-acoustic characterization of flow instabilities on a reduced-scale model of the prototype machine. For this purpose, new electric and hydraulic 1-D models were developed and implemented using SIMSEN, the advanced numerical simulation of systems dynamics tool developed by EPFL.
Extensive experimental campaigns were carried out on reduced-scale models of both a conventional Francis turbine unit and a reversible pump-turbine unit. The main objectives were to, firstly, validate the developed 1-D modelling and to, finally, reach a better understanding of the physical mechanisms responsible for the development of flow instabilities and their interaction with the system. The unsteady mechanical loading of components resulting from off-design operation was addressed, occurrences that can lead to material fatigue and lifespan reduction. The developed methodology was finally validated by performing challenging and ambitious field tests on a full-scale Francis turbine unit. The comparison be-tween predicted and experimental data demonstrated the capability of the methodology in successfully predicting the main features of the hydropower unit dynamic behaviour.
Extensive numerical simulations of the flow through a conventional Francis turbine at various operating conditions and through a reversible pump-turbine in transient operating conditions were performed. It has been demonstrated that the mean characteristics of the flow can be captured, regardless of the complexity of the phenomena. Based on flow simulations, a procedure for the hydro-acoustic characterization of flow instabilities has been developed. Furthermore, a methodology coupling flow simulations and mechanical calculations for the assessment of the dynamic mechanical behaviour of both a conventional Francis turbine and a reversible pump-turbine was validated in the HYPERBOLE project.
The HYPERBOLE project also demonstrated, based on real case studies, the technical benefits hydropower stations can provide for the safe integration of NREs into the existing power network. New power network simulation models for pumped storage power plants with new variable speed technologies were developed and integrated in state-of-the-art power network simulation packages. In addition, new strategies for the exploitation of variable speed technology in electric power markets were envisioned and developed to demonstrate its benefits from an economical point of view.


Project Context and Objectives:
The 20-20-20 strategic energy policy decided by the European Union and its accompanying Renewable Energy Directive is leading towards a dramatic electric energy system transition. The two major pillars of this transition are:
i) a massive penetration of alternative renewable energies;
ii) a broad deployment of energy efficiency initiatives and technologies.
For this goal, hydropower already plays an important role and will increasingly do so to a) contribute to renewable energy production, and to b) provide for the highly dynamic energy storage requirements that enable a widely distributed injection of solar and wind energy into the transmission and distribution systems while preserving their stability through the provision of advanced system services.
The overarching objective of the HYPERBOLE project was to enhance hydropower plant value by extending machine-operating ranges while also improving component long-term availability. More specifically, the project aimed to study the hydraulic, mechanical and electrical dynamics of several hydraulic machine configurations under an extended range of operations: from overload to deep part load. A two-pronged modelling approach relied on numerical simulations as well as reduced-scale physical model tests. Upon suitable concurrence between simulations and reduced-scale physical models results, validation took place on a carefully selected physical hydropower plant properly equipped with monitoring systems. Finally, the benefits resulting from the extended control flexibility provided by a set of hydro units were demonstrated through extensive simulation of the operational conditions of an electric power network with a large variety of sources.
To address this ambitious research plan, a consortium was assembled featuring three leading electric equipment manufacturers of hydraulic turbines, storage pumps, reversible pump-turbines, SME, as well as world-renowned academic institutions. Extensive tests on both experimental rigs and real power plants were performed to validate the obtained methodological and numerical results.

Hydropower plants are key to New Renewable Energy (NRE) integration
In recent years, due to tremendous development and integration of New Renewable Energy (NRE) sources in Europe, hydraulic turbines and pump-turbines have become key to achieving both load balancing and primary and secondary power network control.
To illustrate the challenge of integrating NREs, we can refer to the time history between June 15 and June 30, 2012 of the 3.8 TWh generated in Germany by photovoltaics (PV) power plants and wind farms. The hourly change of the power generation leads to a minimum 10 GW storage capacity, which can only be achieved by hydropower plants. The fast change of power generation by NREs is affecting directly the required operating range of hydro units going from overload down to part load, 20% to 30% of the maximum power, and the number of start-ups and changes between pumping and generating modes, as well. Moreover, only pumped storage power plants are suitable for large-scale electricity storage and fast control over an extended operating range with unrivalled pumping-generating cycle efficiency better than 80 %.
Extreme operating points like deep part load (i.e. with very low discharge) or overload (i.e. with very high discharge) lead the water turbine to experience complex two-phase flow phenomena, i.e. cavitation, which are the source of dynamic loading of turbine components as well as of the complete system including water piping, turbine structure, rotating train, generator, controllers and grid. Moreover, depending on the grid structure, the development of a cavitation vortex can cause unstable operating condition preventing the hydropower plant to be made available to the Transmission System Operator (TSO).

Predicting dynamic loads
Reduced-scale physical model testing (defined by the IEC 60193 standard) are often performed for predicting and assessing the behaviour of prototype machines installed in power plants. In particular, these tests are made to ensure smooth operation, as this may not be reliably predicted by numerical simulations. However, even though a model in similitude with the prototype should behave in the same manner, it is not sufficient for guaranteeing reliable behaviour of the generating units when installed in the hydropower plant. The generating unit behaviour can only be predicted if mathematical models can capture the difference between the test rig hydraulic circuit and the hydropower plant hydraulic system and how they affect the scaled-up hydraulic excitation coming from the machine. In some cases, the induced pressure pulsations may lead to unexpected active power fluctuations coming out of the generator. Therefore, an extensive theoretical investigation needs to follow the reduced-scale physical model testing in order to transpose the reduced-scale physical model test results such as active power, vibration, stress, pressure fluctuations to the real hydropower plant following a methodology developed in the HYPERBOLE project.
To achieve a lean integration of NRE sources within the power network, the TSO needs to be able to quickly perform an energy balance between generation and consumption. Therefore, it is necessary to ensure power plants have enough storage capacity as well as primary and secondary grid control capability to enable this energy balancing process. Both gas fired and hydropower plants are capable of flexible power generation for peaks and regulation ancillary services at a large scale. However, unlike gas-fired facilities, only hydropower plants can exploit renewable primary source energy with almost no emission of greenhouse gas. In addition, pumped storage power plants provide a system to store large amounts of electricity with a full cycle of pumping and generation, which can achieve high efficiency level.
To take on the challenge of enhancing the capability of hydropower plants to be operated over a larger operating range and with a faster response time, the HYPERBOLE Research Consortium partners have defined the three following science and technology objectives to be achieved.

Better understanding of the development of hydrodynamic instabilities and their potential interactions with the hydroelectric system
The first objective is to bring new knowledge of the root causes of operating range instability phenomena for both hydraulic turbines and pump-turbines in order to extend their operating range. When operated at deep part load (below 30 % of rated power) and at overload operation (about 110 % of rated power), hydraulic turbines and pump-turbines experience high levels of vibration and large pressure and power fluctuations. The mechanical components experience high dynamic loads due to the strong and unsteady nature of the flow at these "off-design" conditions. Consequently, the life expectancy of the machine can be significantly reduced, eventually leading to the loss of structural integrity. Moreover, the unit operation may become so unstable that the unit disconnects from the grid. This is the reason why the operating range is typically limited to between approximately 60 % and 100 % of rated power. Additionally, reversible pump-turbines are subject to a dramatic increase in the daily change rate between generating and pumping modes, thereby increasing the time that the machine is exposed to transient conditions that may also endanger the unit integrity. To resolve this first objective, experimental investigations of reduced-scale physical models of both Francis turbine and pump-turbine were carried out at the testing facilities of the project partners. The goal was to identify and extensively investigate the complex flow phenomena corresponding to these sources of excitation by both experimental and advanced numerical tools including Computational Fluid Dynamics (CFD) and Finite Element Method (FEM).

Comprehensive multi-component modelling of hydropower plant dynamics operating in off-design condition
The second objective is to develop comprehensive multi component models of hydropower plant dynamics based on and validated by reduced-scale physical model experiments and extensive field tests on a conventional hydropower plant. The new knowledge achieved for the complex flow phenomena involves the excitation sources of the dynamic loading of the hydraulic turbines when operated at deep part load, part load and overload. This enabled modelling of these sources. These models were implemented for every operating condition case in SIMSEN, the advanced numerical simulation of system dynamics tool developed by EPFL. SIMSEN is widely used by the industrial partners of the research consortium for simulating hydropower plant dynamics, including hydraulic machinery and system, electrical machines, power electronics and the control systems. Advanced dynamic models of the electrical machine and power electronics system were also developed. The resulting comprehensive multi component models of hydropower plant dynamics were validated by reduced-scale physical model experiments and extensive field tests for generating Unit 2 of the MICA ) hydropower plant, located in British Columbia, Canada.

Characterization of the benefits provided by hydropower plants for the integration of NREs into the electrical power network
The third objective aims to demonstrate the benefits hydropower plants can provide to safely integrate New Renewable Energy sources, based on a real selected case study. The enhancement of the operating range of Francis turbines and pump-turbines, as well as the capability of providing inertia emulation functionalities and primary and secondary control capabilities, together with new control strategies will contribute to the increasing use of NREs in power networks. The control flexibility of new hydro units was achieved through the exploitation of advanced energy conversion processes involving power electronic interfaces, either using full-scale converters or advanced excitation systems in Double Fed Induction Generators. As a result, the additional NRE connection capacity was evaluated in a study case with respect to system stability using state of the art power network simulation models. The hydraulic models and related parameters to be considered in the power network simulation were derived by comparison with fully detailed simulation models developed using the SIMSEN simulation software. The detailed simulation takes into account hydraulic waterways, hydraulic machines, electrical machines and the related control systems for hydropower plants. The comparison between detailed and reduced models provides guidelines to model conventional hydro and pumped storage power plants including variable technologies with appropriate level of complexity for stability analysis of power network. In addition to grid stability and security issues of power systems with high percentages of NREs, the HYPERBOLE project aims to develop an optimization algorithm for the identification of adequate strategies regarding the participation of reversible hydropower plants in the secondary reserve market, in complement to participation in the tertiary reserve market.
To achieve these objectives and the overarching goal to enhance hydropower station value by extending the flexibility of the operating range and improving long-term availability, the HYPERBOLE Consortium defined six technical work packages.

Project Results:
Hydraulic excitation sources induced by offdesign operations
As mentioned, hydropower units operating in off-design conditions experience the development of flow instabilities which induce excitation sources for the hydraulic circuit. The resulting flows in the runner inter-blade channels and in the draft tube at the runner outlet are very complex and characterized by cavitation and high levels of turbulence and unsteadiness. For the case of reversible pump-turbines during transient conditions from pump mode to turbine mode, the complexity is reinforced by very fast changes in flow conditions.
Such flows induce additional unsteady mechanical loadings of the components of the machine and can lead to the reduction of their lifespan. One way to investigate such flow conditions and the resulting excitation sources is to perform CFD (Computational Fluid Dynamics) flow simulations in the full machine. However, these simulations are quite challenging due to the complexity of both geometry and flow. Another way is to perform experimental measurements on a reduced-scale physical model in a laboratory. Both approaches have been carried out in the framework of the HYPERBOLE project. They are described below, together with the corresponding milestones and results.

Unsteady flow simulations of a Francis turbine at full load, part load and deep part load
Unsteady two-phase flow simulations have been carried out on a reduced-scale physical model Francis turbine under deep part load, part load and full load conditions. Each operating condition is characterized by unique cavitation phenomena and flow structures. This required the development of adapted simulation approaches for the different operating conditions. The three operating points in focus were investigated using a commercial flow solver package (ANSYS CFX). Evaporation and condensation have been accounted for by employing a multiphase and cavitation model. Results from the experimental investigations of the model turbine were used to prescribe adequate boundary conditions for the simulations. The mass flow rate or alternatively a time varying total pressure was prescript at the inflow. To ensure a comparable cavitation behaviour an adequate pressure level was set at the draft tube outflow that is described by a similarity number (Thoma number).
At deep part load condition, two operating points with different runner rotational speeds have been investigated. Clearly visible inter-blade cavitation vortices inside the blade channels were observed on the test rig for higher runner speeds. By decreasing the runner rotational speed these vortices disappeared almost entirely, and when they were visible they occurred only intermittently and were scattered over the different channels. Regardless of what was physically visible, stochastic pressure fluctuations were measured in the draft tube cone and on the runner blades at all deep part load operating points. These stochastic fluctuations are induced by strong turbulence and flow separations. These kinds of flow conditions along with turbulence induced cavitation are extremely challenging for numerical flow simulations in terms of accuracy and stability. Two equation turbulence models, as used for standard hydraulic design work, were found to be inapplicable. By employing so called second generation turbulence models, good correlations with the measured pressure fluctuations have been achieved. Pre-studies investigating the influence of space and time discretization on the resolution of the inter blade vortices demonstrated the need for small cell sizes, especially in the runner domain using short time steps of around 1° runner revolution/time step. Although the short time step allowed for the analysis of the loads resulting from the rotor-stator interaction and a detailed resolution of the vortex structures, this placed high demands on the computational resources, since simulation times of multiple runner revolutions were required to ensure a satisfying frequency resolution.
For the part load conditions, two operating points at 80% and 64% of the discharge at the Best Efficiency Point (BEP) have been numerically investigated. At part load, the flow exiting the runner has a significant tangential velocity component. This swirling flow generates a helical shaped vortex rope in the draft tube cone, which rotates with the Rheingans frequency underneath the runner and generates periodic pressure oscillations. At a lower pressure level, this vortex rope is filled with a large amount of vapour. As the measurements showed, the quantity of vapour has a tremendous effect on the pressure oscillations due to a shift of the eigenfrequencies of the test rig, which in the worst case implies resonance. The simulations were performed at resonance free conditions to satisfy the constant mass flow boundary condition. For the characteristic Rheingans frequency, a very good match with the measured values has been achieved. A reduced numerical model without the spiral case and only one distributor passage was found to be sufficient to resolve the vortex rope and the corresponding rotational frequency even with large time steps, but the effects resulting from the rotor stator interaction are neglected in this way. Further investigations concentrated on the in-depth modelling of the vortex structures at lower pressure levels. Mesh refinement, smaller time steps and the use of a scale resolving turbulence model improved the resolution of the vortex structures and enhanced the correlation with the high-speed visualizations.
At full load condition, the increasing flow rate induces an increasing counter rotating swirl exiting the runner and thus entering the draft tube. Inside the draft tube an axial symmetric vortex rope is developing, which promotes large cavitation development. A massive self-excited pressure surge may arise, depending on the eigenfrequencies of the hydraulic system. Since the hydraulic system is not included in the 3D flow simulation, a fixed mass flow condition fails to predict the self-excitation. Extensive studies investigating different inlet boundary conditions with time varying sinusoidal excitations have been conducted with the goal to identify appropriate inlet conditions independent from measuring data. The investigations showed that by using time varying mass flow or total pressure boundary conditions, any behaviour could be generated. For an optimal fit of the observed pressure surge on the test rig, the measured pressure fluctuations at the spiral case were extracted, smoothed and applied as a total pressure condition at the spiral case inlet. In this way, the simulation is capable of reproducing the behaviour measured on the test rig. The simulation predicts cavitation sheets on the suction side of the runner blades whose size oscillate synchronously with the pressure fluctuations in the cone. However, the small bubbles exiting the runner domain after the vortex core collapse as observed on the test rig could not be resolved by the simulation.
For all operating conditions, a fine qualitative correlation of the simulated cavitation regions were found. Therefore, isosurfaces of the vapour volume fraction were compared with video sequences gained from high speed video recordings. The simulations were able to predict successfully the impact of pressure level changes on the pressure pulsation and cavitation phenomena. The measured velocity profiles, extracted from Particle Image Velocimetry (PIV) and Laser Doppler Velocimetry (LDV) measurements, were in good agreement with the simulated time averaged velocity profiles. Larger deviations to the measurements were found only for the predicted head, especially at very low discharge. In order to reduce the simulation complexity, geometry simplifications, such as the neglect of the runner side gaps and runner crown bores, were accepted. However, this reduces the pressure loss of the runner and provides an explanation for the observed deviations. The measured pressure signals at the guide vanes, the runner blades and in the draft tube cone were thoroughly compared with the simulation results in time and frequency domains. The comparison for the on-blade sensor SS_TE1 located on the suction side near the crown for the different operating regimes is shown in Figure 6. The gained pressure fields on the runner were used for a quasi-stationary finite element analysis in the framework of Work Package 2.

Unsteady flow simulation of a pump-turbine during transient operating conditions
Unsteady flow simulations of a model-scale pump-turbine have been carried out for the transition from pump mode to generating mode within 8 seconds, assuming a linear variation of the runner rotational speed corresponding to a mode change time of 32 seconds at prototype scale. The results of two different codes have been evaluated, a commercial software and an open-source code.
Under these transient conditions, the pressure difference between turbine inlet and turbine outlet changes over time in a real power plant, which is reproduced in the model by using a suitable test setup. As the experimental flow rate serves as input for simulation, this change in head is a result of the simulation and can be compared to measurement. The first observation is that the codes agree extremely well for pump mode and the beginning of pump brake mode, when the runner is still rotating in pump direction but flow direction has reversed. Later in pump brake mode, some small deviations are found. In generating mode, the commercial software systematically predicts slightly higher values than the open-source code.
For the open-source code, a finer mesh gives comparable results for head in pump mode and pump brake mode, and a lower head in generating mode. Although it is known that the numerical models used in this study have their limitations in strongly stochastic turbulent flow conditions, a good agreement is found between simulated and measured values of head. Two of the key findings are: that an open-source code gives comparable results to a commercial code, and that a fair first estimate of the flow behaviour can be obtained with a coarse mesh.
The off-design operation during the transient leads to pressure fluctuations in the machine that are thoroughly investigated. By comparing the time signals, short-time Fast Fourier Transforms (FFT) and wavelet transforms to identify the frequencies that might excite the structure, it is found that for pressure sensors at different locations in the machine, the simulation is capable of predicting the general slope of the curves, but has difficulties in identifying the frequencies and amplitudes.
The most remarkable difference between the codes is found in the draft tube. Both simulations predict an asymmetric rotating vortex rope as shown in Figure 7, but this is found to be significantly more pronounced in the commercial code. However, even the commercial code does not reflect the full extent of the pressure fluctuations from the experiment, where the fluctuations start approximately 2 seconds earlier and are more pronounced in amplitude.

Francis turbine reduced-scale model test
For the first two objectives of the HYPERBOLE project, and for analyses and modelling tasks within different work packages of the project, a comprehensive and reliable experimental database of the reduced-scale physical model of MICA hydropower plant Unit 2 Francis turbine was collected through three measurement campaigns in the laboratory of hydraulic machines at the EPFL, Switzerland. The prototype homologous reduced-scale model turbine was equipped for various tests with different measurement devices to analyse the flow field in multiple regions of the machine under different operation conditions in order to define the overall machine behaviour over a wide operating range. The second project objective required special instrumentation of the test rig piping system, including the implementation of a variable speed pump for periodic flow excitation, in order to perform a thorough modal analysis to determine the hydroa-coustic eigenfrequencies of the test rig.
The complete operating range of the reduced-scale model turbine was analysed with a particular focus on deep part load (DPL), part load (PL) and full load (FL) operating conditions using special measurement techniques. The selected measuring equipment was applied to improve the understanding about physical phenomena limiting the safe operating range of the machine.
Optical methods such as Laser Doppler Velocimetry (LDV) and Particle Image Velocimetry (PIV) were applied for the determination of unsteady velocity fields in the draft tube cone in cavitation free conditions and under the presence of a self-oscillating vortex rope at full load and a rotating helical shaped vortex rope at part load. With high-speed visualization systems, the cavitation phenomena developing below the runner in the draft tube cone was visualized in full load, part load and deep part load condition. In addition, the runner blade channel flow at deep part load was visualized through the draft tube cone. To analyse the behaviour of the inter-blade cavitation vortices at deep part load and to have a better database for correlation and model development, a novel technique of flow visualization from within the guide vane channel was set up using a special camera system embedded into the guide vanes as shown in Figure 8 (right).
The model machine was instrumented with pressure transducers for pressure fluctuation measurements at selected locations to analyse critical instabilities with high levels of vibrations and large torque and pressure fluctuations.
During the experimental campaigns, the optical flow observations by LDV/PIV and high speed visualisation and measurement of pressure and torque fluctuations were performed together with the standard parameter monitoring of the model machine. The head, discharge, runner rotational speed and Thoma cavitation number were recorded for the operating conditions analysed by numerical methods (Computational Fluid Dynamics – CFD and Finite Element Analysis FEA) and expected during the onsite tests at the homologous MICA prototype machine.
The visualization of the cavitation vortex rope development in the draft tube cone from full load to part load is presented in Figure 9 (upper row) for various operation points. In addition, the structure of the cavitation vortex core of the cavitation inter-blades vortices observed through the guide vane at deep part load condition (DPL2) is shown.
Selected numerical results for the same operating points are shown in the lower row of Figure 9. The numerically predicted cavitation volume is displayed by an isosurface of the vapour volume below and within the runner and in the draft tube. The results of the numerical flow simulations are in good agreement with the experimental visualisations.
The results gained during experimental campaigns on the reduced-scale model were used in further work packages for CFD validation and hydroacoustic characterization of flow instabilities. They showed a high quality level with respect to data density and resolution of the observed and investigated instationary hydraulic phenomena.

Hydromechanical test on model turbine
The second experimental campaign on the reduced-scale model machine was intended to gain a hydro-mechanical characterization of the reduced-scale model operating in off-design conditions and to provide a database for validation of CFD and FEA results. Another objective was to gain correlation data for the field measurements on MICA hydropower plant Unit 2.
To achieve these objectives, a similitude between model and prototype with respect to hydraulic and structural mechanical behaviour had to be fulfilled. For this purpose, an instrumented runner was manufactured and onboard instrumentation was installed on the runner blades that included 24 strain gauges and 6 pressure transducers. The challenge of the second measurement campaign lay on the runner manufacturing, the integration of the instrumentation and data transfer from an onboard system via telemetry to the data acquisition system on the test rig.
The mechanical homology requirement on the runner led to a monobloc runner manufacturing procedure. In order to minimize the impact of cabling and sensors on the hydraulic behaviour of the runner, the cables and sensors were directly integrated into the blades. This integration minimized the impact on the structural behaviour of the runner by applying minimized sensors, which are still able to resolve pressures and forces acting on the runner. The integration of strain gauge sensors into one runner blade is shown in Figure 10. The same principle is used for the pressure transducers installed on the runner blades. The CAD model is shown in Figure 10a, the grooves and sensor accommodations in Figure 10b and Figure 10c, respectively, and the coating of sensors and cabling in Figure 10d. The complete monobloc runner is shown in Figure 10d, together with the visible cables and connectors used for contact to the telemetry and data acquisition system.
The results from the instrumented runner were used for validation of numerical CFD analyses and structural mechanical evaluations. The onsite measurements campaign also used the post-processed data of the reduced-scale model tests.

Mechanical structure dynamics
In off-design conditions, the mechanical structure of both Francis turbines and reversible pump-turbines experience high dynamic loads due to the unsteady nature of the flow. Consequently, the life expectancy of the machine can be significantly reduced, eventually leading to the loss of structural integrity. This is the reason why the operating range is typically limited to between approximately 60 % and 100 % of rated power. Additionally, reversible pump-turbines are subject to a dramatic increase in the daily change rate between generating and pumping modes, which also may endanger the unit integrity.
In the HYPERBOLE project, the partner GE Renewable Energy has developed and validated, in collaboration with the other partners of the consortium, a methodology for the calculation of the dynamic mechanical behaviour of both Francis turbines and reversible pump-turbines in various operating modes, including full load, part load, deep part load and transient modes. The validation was done by comparison of numerical results with experimental measurements on both the reduced scale physical model and the full-scale turbine. Experimental tests have enabled a better understanding of the impact of the hydrodynamic instabilities occurring in off-design conditions on the mechanical loading of the runner.
The developed methodology is a decisive step towards both the transposition of the mechanical dynamic behaviour from model scale to prototype scale and an assessment of the lifespan of the mechanical components of the machine by taking into account the operating time in off-design conditions.

Unsteady behaviour of Francis turbines in off-design conditions
The unsteady mechanical behaviour of a Francis turbine runner operating in off-design conditions has been widely investigated by means of both experimental tests and numerical simulations. The general procedure followed for this task is summarized in Figure 11. As shown, experimental tests performed on both the reduced scale physical model and the full-scale turbine have enabled the validation of numerical simulations by Finite Elements Analysis (FEA). In addition, experimental results have been transposed to the prototype scale and compared with the onsite measurements performed on the full-scale turbine.
Firstly, the static mechanical behaviour of the full-scale runner of the MICA Francis turbine has been numerically studied at the nominal operating point for which the runner was designed. In addition, vibration mode shapes of the full-scale runner were studied in order to understand how the runner can respond to a dynamic solicitation that could induce an amplification of the runner deformation if the excitation frequencies is close to the runner eigenfrequencies. The shape of the modes obtained by numerical simulation have been validated by comparison with the ones measured during the experimental tests on the MICA full-scale turbine.
An important part of the study focused on the understanding of the unsteady mechanical behaviour of a Francis turbine runner under three operating conditions: full load, part load and deep part load. The impact of the axisymmetric vortex rope at full load, the helical vortex rope at part load and the Rotor-Stator Interaction (RSI) were identified within the operating range of the machine and investigated in detail. The impact of hydroacoustic resonances induced by the axial rope at full load and the helical vortex rope at part load has also been investigated on both the reduced scale model and the full-scale turbine. As presented previously, these phenomena have been modelled by unsteady Computational Fluid Dynamics (CFD) calculations and the simulated pressure fields on the runner blades have been used as input for the mechanical simulations in order to simulate the runner mechanical unsteady behaviour under those different phenomena. Quasi-static assumption has been considered for FEA since the frequency of the excitation induced by the hydraulic phenomena is significantly lower than the eigenfrequencies of the reduced scale runner and the risk of structural dynamic amplification is therefore prevented. The dynamic stress distribution on the runner blade of the reduced scale model of a Francis turbine obtained for three different operating conditions is given in Figure 12.
In addition, mechanical tests have been performed with a reduced scale runner in partial similarity with the full-scale turbine in order to improve the mechanical homology with the real runner. The numerical results have been compared to the experimental ones measured on a test rig with the reduced scale model. Globally, the results of the numerical simulations are in good agreement with the measurements as shown in Figure 13. This demonstrates the capability to simulate correctly the different hydraulic phenomena. The simulation enables the prediction of the mechanical response on the entire runner and not only on discrete locations such as is the case for the model tests. Finally, measurements have been made on the full-scale turbine of MICA hydropower plant. The measurements aimed to validate the transposition rules between the reduced scale model and the prototype. This aspect is crucial since the objective was to verify the possibility of predicting the mechanical behaviour of a full-scale turbine by performing only measurements on a reduced physical scale model. Due to differences between the reduced scale model and the full-scale runner, mainly in the locations of the strain gauges as well as in the geometries and the tested different operating conditions (the conditions are not fully controllable on the full-scale machine), the verification of the transposition rules with a high confidence level was not possible. Additional mechanical calculations on the full-scale runner could help to estimate the impact of the differences between model and prototype runners on the transposition rules validation.

Dynamic behaviour of reversible pump-turbines in transient mode
The mechanical dynamic behaviour of a reversible pump-turbine during mode transitions has been investigated by numerical simulations and experimental measurements. A pump to turbine transition has been experimentally performed on a reduced scale model in the laboratory of the partner Andritz Hydro GmbH. The objectives were to study the feasibility of simulating the mechanical behaviour during transient mode operation, when the turbine experiences the development of a very complex flow in order to evaluate the impact of the pump to turbine transition on the machine mechanical structures.
Unsteady CFD calculations have been performed by the partner USTUTT in order to simulate the different hydraulic phenomena occurring during such a transition. Comparisons between the CFD calculations and the pressure measurements show that a part of the dynamic hydraulic excitation is not yet well captured by the numerical simulations. This type of simulation was expected to be very complex and requires many calculations with a very small time-step.
The mechanical simulation has been performed on a trial basis to evaluate its feasibility. Such a complex mechanical calculation was challenging. A transient simulation (which is the most complex analysis in mechanical simulation) considering the fluid volumes surrounding the structure with a large number of iterations (about 1000) was performed. If required, the mechanical calculations can be improved by using more iterations and a smaller time step, closer to the values used for CFD simulations. The mechanical response of the runner was then compared to the measurements made on the reduced scale model.
As expected, the mechanical simulation, which is dependent on the CFD calculation, cannot exactly reproduce the mechanical dynamic behaviour of the runner since the hydraulic loadings simulated by CFD do not match completely with the experimental results.
However, the numerical mechanical calculation allows drawing a runner mechanical behaviour coherent with the experimental one. Such transient simulation can be considered as a new milestone for the understanding of key phenomena occurring during transient operation and for the improvement of the reactivity of the pump turbine during mode shifts.

Electrical machine and power electronics
As already mentioned in Section 2, electrical power systems of the future will need more flexible generating units to compensate for the non-predictable generating behaviour of New Renewable Energy (NRE) sources. In that respect, more flexibility in terms of operation range and response time is required. Classical hydropower generation is limited in both requirements, but the ongoing developments in power electronics now offers hydropower operation concepts with variable speed to provide higher flexibility.
The high contribution of NREs in the electrical power generation mix results also in missing grid inertia, so the power system could react with impermissible frequency deviations in case of loss of generation due to a system fault.
By using frequency converters – offering a wide operation range and a short response time – future hydro generation systems can actively contribute to electrical power stability. From an electrical point of view, two options for speed variation of generation units are possible:
- Double Fed Induction Machine (DFIM): Here an electrical machine with a full 3~ winding system in the rotor is used. The Rotor circuit can then be connected by a frequency converter to the grid;
- Full Size Frequency Converter (FSFC): Here a converter with 100% of the unit size is used. The electrical machine is a classical Synchronous Machine, also used in fixed speed power generation systems.
Both options were investigated in the HYPERBOLE project by the partner Andritz. The topics that are described in the following are:
- Possible Operating range;
- Consideration of losses / efficiency;
- Transient operation / Response time;
- Power & Speed control concepts;
- Fulfilment of the requirements from the Network Code.
Operating range
The requested operation range is defined by the hydraulic machine and the operating conditions (requested hydraulic head and power variation). For the DFIM concept, the converter size is defined by the requested speed range. FSFC systems do not have this kind of limitation.
System Losses and Efficiency
For the evaluation of the electrical system losses, the layout of a 420MVA unit was used (see Figure 14). Due to 100% power conversion, the efficiency of a FSFC solution is a little bit lower compared to DFIM. This drawback is compensated by higher system flexibility. The preferred concept must be evaluated case by case.
Transient operation / Response time
Power electronics with an appropriate control strategy offer response time for power control in tenths of a second range. This feature allows the use of a more sophisticated control algorithm to implement fast voltage control in the grid and fast grid power control of the unit.
Power & Speed control concepts
Due to an additional degree of freedom with variable speed concepts, power to the grid and speed of the turbine can be controlled. Shorter response time can be achieved by using the converter for power control and the hydro governor to control the speed of the unit. This approach also provides “synthetic or virtual” inertia to the grid.
Fulfilment of the requirements from the Network Code
A mandatory requirement from the Transmission System Operator (TSO) is to fulfil the so-called Network Code requirements for connecting the generating unit to the grid. The most problematic case for converter operated units is the Fault Ride Through (FRT) which demands that the units must stay connected to the system during a defined voltage dip on the grid connection point. Here the DFIM solution shows its drawback that high rotor currents must be handled during and after the fault. Here the FSFC has the advantage that the current can be controlled quickly and accurately.
In summary, the conclusions of the investigation are as follows:
The concept using DFIM has disadvantages in terms of optimal flexibility and higher technical effort for design of the generator;
The solution using FSFC with a classical synchronous machine has a higher capital expense, but offers more technical features in plant operation (better transient performance, opportunity for fast grid voltage control and fast transition time between turbine and pump operation).
These investigations were building the bases for the grid integration studies worked out in WP6 “Power Plant Network Integration”.

Reversible pump-turbine model transient mode change test
A drastic reduction in the time required for changing operating modes of reversible pump-turbines is one key element to providing improved services for the integration of New Renewable Energy sources and was investigated in the context of S&T objective 3. A fast transition from pump to turbine operation was identified as a mode change of particularly high importance since it allows the injection of large amounts of power into the grid in a very short time.
The characteristics of flow change during such a transition is relatively similar to a load rejection in pumping mode, something that is normally considered a rare case in the lifetime assessment of a pump-turbine. However, regular usage of such fast transitions on purpose requires a better understanding of flow behaviour, assessment of structural loads on the machine components as well as validation of the simulation methods used for assessment of the safety of the entire hydropower plant regarding water hammer and other 1D system analysis.
For this purpose, a discovery experiment by modification of an existing 4 quadrant pump-turbine model test facility was implemented. Such test facilities are typically designed to guarantee as stable as possible operation. From test rig control software to data acquisition, all elements of the test rig are designed to provide reproducible time averaged values for specified steady state operation. Therefore, modifying such a test rig to produce and measure fast transitions between pump and turbine operation poses numerous technical challenges.
The retained solution, depicted in Figure 16, was to replace one of the service pumps of the test rig by a pipe section and a tailor made throttle plate. The control system was also modified such that both the pump-turbine model and the remaining service pump could be operated with variable rotation rates. For the considered discovery experiment, linear ramp functions were proven sufficient to implement the operation mode changes in such a way that they corresponded to the possible operation of a large scale power plant. The standard time-averaging flow meter was replaced by a high frequency magneto-inductive flow meter. All time-averaging pressure measurement equipment were seconded with dynamic pressure sensors.
Transitions from turbine to pump mode and vice versa are measured for transition times of 4s, 8s and 20s. A transition time of 8s at model scale is transposed to a time of about 30 s for a large scale hydropower plant. A transition time of 4s shows effects of a fast transition that would only be realistic on relatively small power plants, whereas transition time of 20s is close to a quasi-steady transition. Figure 17 shows the prescribed rotation rates of the service pump and the pump-turbine scale model and the resulting operation conditions for a transition time of 8s. The fast reversal of flow rate and the characteristic drop in head during the transition periods is measured as predicted by 1D modelling before the test.
The pump turbine model is instrumented with pressure sensors and strain gauges on the runner, and pressure sensors in the head cover between the guide vanes and in the draft tube. Measurement time series of these sensors were provided for comparison with 1D system modelling, flow and mechanical simulations. While the experiment fully confirms the validity of common 1D simulation methodology using steady state 4quadrants characteristics, it shows that state of the art flow simulations cannot resolve all phenomena encountered accurately.
The direct analysis of pressure fluctuations measured during fast to slow transition, as shown in Figure 18, reveals that, regarding the pressure fluctuations on the runner as well between the guide vanes, the fluctuations encountered during a slow transition are more severe than when traversing the unfavourable flow regimes of pump brake mode (PUBR) as fast as possible. It also shows that for the study of such highly dynamic flow regimes, series of multiple quasi-steady-state considerations or measurements can safely be used as a worst case assessment for dynamic loads in the pump-turbine. However, a transient test, once set up, allows for the reproduction of the flow regime change in a realistic manner in a short time.

Field test based validation
The development of methodologies enabling the prediction of the dynamic behaviour of hydropower plants required validation by direct measurements on a full-scale turbine. To meet this need, the partner UPC performed in collaboration with other consortium partners extensive field tests of the MICA hydropower plant Unit 2 in British Columbia, Canada. This unit has a rated power of 444 MW and a runner diameter of 5.4 meters. Before the final field tests, a remote online monitoring system was installed. This monitoring system was recording the operating parameters of the generating unit over one year before the final tests. The objective was to optimize the final test strategy, helping to select the best position for test sensors and to determine output power values to be tested. The final field tests were carried out in November 2016 and involved the partners UPC, Voith Hydro, GE Renewable Energy and EPFL in collaboration with BC Hydro (BCH), the owner of the power plant. An important number of sensors was installed in both the stationary and rotating parts of the machine to validate numerical and experimental tests on the reduced scale model and to analyse the dynamic behaviour of the prototype.

Installation of a remote monitoring system in MICA hydropower plant
An online monitoring system was installed in MICA hydropower plant Unit 2. It included vibration sensors, pressure sensors and operating signals of the prototype. In total, 20 signals were recorded over one complete year. More specifically, two pressure sensors were installed in the spiral casing and the draft tube. Moreover, eight accelerometers were installed in different parts of the machine: two radially in the turbine bearing, two radially in the generator bearing, two axially in the thrust bearing, one in the spiral casing wall and another in the draft tube wall. In addition, the signals of four displacement probes that were already installed in the machine were also monitored. The operating parameters such as the wicket gate opening, the rotating speed of the runner and the active power were also recorded. These parameters were provided by the SCADA system already installed at the power plant.
All the sensors were connected to an acquisition system installed near the control room of MICA hydropower plant Unit 2. This acquisition system was configured to measure all the aforementioned parameters every hour while the machine was operating. The database was accessible using a remote VPN connection. The monitoring of MICA hydropower plant over one complete year enabled the collection of an important database. This allowed the HYPERBOLE partners to analyse the behaviour of the machine over one complete year and to prepare the onsite measurements held in November 2016. Moreover, the additional data recorded by the monitoring system provided additional validation of the empirical laws established on the reduced scale model.

Onsite measurements at the MICA hydropower plant
In November 2016, the partner UPC in collaboration with the partners Voith Hydro, GE Renewable Energy and EPFL performed measurements of the MICA hydropower plant Unit 2. The work, including the installation of the equipment and the measurements, was realized in collaboration with BC Hydro who gave access to the unit and provided logistical and technical assistance. Due to the size of the machine, the number of sensors to be installed and the number of tests to be realized, the task was ambitious and challenging.
The measurements aimed at studying the mechanical and hydrodynamic behaviour of the machine when operating under various operating conditions, including deep part load, part load and full load, transient operations, synchronous condenser mode and load rejection. Moreover, the collected data have allowed the HYPERBOLE consortium to validate the developed methodology for the transposition of the hydro-mechanical behaviour of Francis turbines from the model scale to the prototype scale. In addition, the injection of air through the head cover of the turbine was also tested in order to investigate its impact on the hydrodynamic instabilities experienced by the machine in offdesign conditions, including hydro-acoustic resonance at part load and self-excited behaviour at full load.
For this purpose, sensors were installed in the rotating and stationary parts of the machine. The runner blades were instrumented with 24 strain gauges and 8 pressure sensors. In the stationary parts, three pressure sensors were installed downstream of the runner in the draft tube cone and five sensors were installed upstream of the runner, including three sensors in the spiral casing and two sensors in the penstock. Moreover, the operating parameters of the machine, i.e. the wicket gate opening, the rotating speed of the runner and the active power, were recorded to properly identify the investigated operating points and to facilitate the comparison with the measurements performed on the reduced scale physical model. The mechanical torque on the shaft was also measured using strain gauges. In total, 64 channels were acquired simultaneously using a distributed acquisition system. A view of the different signals recorded at different locations of the machine is given in Figure 21 when the unit was operated in full load conditions.
Different operating conditions were investigated, from very low load conditions (P = 34 MW = 8% of Prated) to full load conditions (P = 475 MW = 107% of Prated). At part load conditions, a hydro-acoustic resonance was identified at an output power of about 250 MW, i.e. 56% of Prated. Resonance conditions are induced by a matching between the precession frequency of the vortex, which acts as an excitation for the hydraulic circuit, and one of the eigenfrequencies of the hydraulic circuit. As shown in Figure 22, the amplitude of the pressure fluctuations is severely amplified in such conditions. However, it has been observed that the occurrence of part load resonance does not affect the mechanical loading of the runner, confirming the observations made on the reduced scale physical model. The frequency of the pressure pulsations in resonance conditions was different by 10 % of the value predicted by 1D hydro-acoustic models developed in the HYPERBOLE project. Despite the slight discrepancy, this result is an important milestone since it demonstrates the ability of the developed models to predict the frequencies of interest with an acceptable uncertainty.
At full load conditions, a self-excited instability was observed at an output power of about 475MW, i.e. 107% of Prated. It is characterized by an amplification of both the pressure fluctuations in the hydraulic circuit and the unsteady mechanical stresses on the runner blades. In addition, torque fluctuations are also magnified, resulting in output power fluctuations with a peak-to-peak amplitude of about 30 MW, i.e. 6% of the average power.
The frequency of the instability observed on the prototype was in agreement with the value predicted by 1D hydro-acoustic models, which demonstrates the capability to predict the frequency of the instability.
Finally, it has been demonstrated that the injection of air does not affect the instabilities occurring in both part load and full conditions, contrary to the observations made on the reduced scale physical model. This can be explained by an inadequate quantity of air injected into the machine, which is not controllable by the operator on the prototype contrary to the reduced scale model.

Hydroelectric system modelling
In off-design conditions, the development of cavitation phenomena in the draft tube of Francis turbines modifies drastically the dynamic behaviour of the machine. For instance, it introduces an additional compressibility in the draft tube, resulting in a drastic change in the value of the eigenfrequencies of the hydraulic circuit. Moreover, such flow induces excitation sources for the hydraulic circuit, inducing pressure fluctuations in the whole system and finally output power pulsations.
Experimental tests are commonly performed on reduced scale physical models enabling the prediction of the hydraulic behaviour of the prototype machine in terms of efficiency and cavitation with a good confidence level. However, although both geometric homology and kinematic similarity between model and prototype scales are fulfilled according to the IEC 60193 standard, the amplitude of pressure fluctuations cannot be directly transposed from the model to the prototype scale as the eigenfrequencies of the system depend on both the cavitation volume and the characteristics of the hydraulic circuit.
In this context, the HYPERBOLE project has implemented a novel methodology for predicting the dynamic behaviour of hydropower plant units. It is based on the experimental and numerical characterization of the hydrodynamic instabilities experienced by the machine on a reduced scale model in laboratory and the development of a proper one-dimensional (1D) model of the full hydroelectric system, including the cavitation flow in the draft tube of the machine. This methodology, which has been finally validated by comparison with experimental results measured on the full-scale turbine, is illustrated in Figure 23.
Understanding of the physical mechanisms responsible for the hydrodynamic instabilities development
The root causes of the hydrodynamic instabilities experienced by Francis turbine units are not fully understood and need to be investigated in order to propose 1D modelling that respects the involved physical mechanisms. This has been made possible within the HYPERBOLE project by extensive experimental measurements performed at EPFL on the reduced scale physical model of MICA hydropower plant Unit 2, which included the investigation of:
- the selfexcited cavitation vortex rope in a Francis turbine draft tube and its interaction with the hydromechanical system at full load condition;
- the precessing vortex rope in a Francis turbine draft tube and its interaction with the hydraulic circuit at part load conditions;
- the interblade cavitation vortices in a Francis turbine runner and their mechanical impact.
Among the main findings, the role of the coupling between the draft tube flow, the cavitation sheet developing on the runner blades and the self-excited vortex rope at full load has been demonstrated. The periodic appearance and disappearance of the cavitation sheet on the runner blades plays a crucial role in the torque fluctuations experienced by the shaft of the turbine and the corresponding power fluctuations. These results pave the way to a full understanding of the onset of the self-excited instability at full load and its interaction with the hydro-mechanical system that leads to unacceptable power fluctuations and compromises the extension of the operating range of the machine. Notably, a more accurate modelling of the dynamic behaviour of the hydroelectric system in such conditions would enable the accurate prediction of unstable conditions.
At part load conditions, the precession of the vortex rope acts as a pressure excitation source for the hydraulic circuit, inducing pressure pulsations that are strongly amplified in case of resonance with the hydraulic circuit. It is therefore necessary to accurately identify the value of the precession frequency and the natural frequency of the system on the complete operating range of the machine at part load conditions, including variations of head and discharge. This however requires a wide number of operating point measurements on the reduced scale physical model of the turbine, which is unacceptable in terms of time allocation for an industrial project. The HYPERBOLE project has demonstrated that the influence of both the head and the discharge on the precession and natural frequencies can be represented by a single parameter, the swirl number. Concretely, this means that measurements for only a limited number of values of discharge at a given head value are sufficient to predict both frequencies in the complete part load operating range.
The investigation of inter-blade cavitation vortices observed at deep part load conditions is still in its early stages and the HYPERBOLE project was one of the precursors. A novel experimental device for visualizing the development of inter-blades cavitation vortices has been developed in the framework of the HYPERBOLE project. By using CFD simulations, it has been demonstrated that the development of inter-blade vortices is related to the development of backflow near the runner hub. Moreover, it has been highlighted that the head of the turbine plays a crucial role in the formation of cavitation inter-blade vortices: they are developing only for a given range of head values.

Modelling of the hydroelectric system for the reduced-scale physical model and the full-scale turbine
The dynamic behaviour of a hydropower plant can be predicted for various operating conditions by using one-dimensional (1D) hydro-acoustic models. The geometrical characteristics of the hydraulic circuit must be perfectly modelled, but this relies on parameters that are well known, such as the length and thickness of the pipes and their elasticity. However, when hydropower plants operate in off-design conditions, the development of cavitation flow in the draft tube of the machine strongly impacts the eigenfrequencies of the hydraulic circuit, as it increases the flow compressibility in the draft tube. Therefore, a set of parameters representing the local momentum excitation source and the additional compressibility are introduced in the 1D model of the hydroelectric system, such as the cavitation compliance. A detailed modelling of the hydraulic circuit and the cavitation draft tube flow at model and prototype scales has been developed in the framework of the HYPERBOLE project by using the SIMSEN software. Additional parameters have been implemented in the SIMSEN software to take into account additional physical mechanisms, such as the dissipation induced by the periodic phase change in the draft tube.
In addition, new electrical models for simulating low load torque fluctuations developed in the framework of the HYPERBOLE project have been integrated into the SIMSEN software and applied to the test cases investigated in the HYPERBOLE project, i.e. MICA and Aldeadàvila II hydropower plants.

Hydro-acoustic characterization of hydrodynamic instabilities on the reduced-scale physical model
The hydro-acoustic parameters modelling the cavitation draft tube flow strongly depend on the operating conditions and need to be carefully calibrated at the model scale before being transposed to the full-scale machine. One way to achieve this calibration is to identify experimentally the eigenfrequencies of the hydraulic circuit by using an external excitation system, such as a rotating vane injecting a periodical discharge at a given frequency in the circuit. The hydro-acoustic parameters can then be identified based on the value of the eigenfrequencies of the system and a proper 1D modelling of the hydraulic circuit. Such a methodology has been developed in the framework of the HYPERBOLE project and could lead to the revision of the scale-up relating to oscillating phenomena in the IEC 60193 standard for industrial model testing in the future.
Another way to achieve the calibration of these parameters is to carry out 3D simulations of the flow through a Francis turbine at unstable operating conditions (under resonant conditions). 3D simulations are able to capture the vortex downstream the runner with the development of cavitation inside the vortex. Then, based on the results provided by the 3D simulations, an identification procedure has been put in place allowing the calibration of the parameters used in the 1D models.

Transposition from reduced-scale physical model to full-scale turbine and validation by onsite tests
The hydro-acoustic parameters used to model the cavitation draft tube flow were identified at the model scale for several operating points. It has been demonstrated that it is possible to predict them over the complete operating range of the machine by expressing their value as a function of the swirl number, which is a major result for the 1D modelling of hydropower plants.
The corresponding empirical laws have been transposed to the prototype scale by using scale-up laws developed and validated in the framework of the HYPERBOLE project. By injecting the obtained values in the 1D model of the corresponding hydropower plant, it was possible to compute numerically its natural frequencies and predict the resonance conditions over the complete part load operating range of the prototype. At full load, the frequency of the instability, which corresponds to the first eigenfrequency of the system, has been determined over the complete full load operating range with the same procedure.
The numerical results have been compared with the experimental results measured in the MICA hydropower plant Unit 2 (British Columbia, Canada). As illustrated in Figure 25, a resonance excited by the part load vortex rope was experimentally identified in the real machine at an output power comprised between 57% and 59% of the rated output power of the machine, whereas the numerical modelling predicted resonance conditions at an output power equal to 65% ± 2% of the rated output power. The discrepancy between the experimental and numerical results is comprised between 4% and 10% of the rated output power of the machine. This result represents a spectacular advancement for the accurate assessment of hydropower plants stability as it demonstrates that the developed methodology and the associate 1D modelling are able to capture the resonance conditions on the prototype with an acceptable uncertainty.
In terms of frequency, the empirical laws established on the reduced scale physical model for the precession frequency of the vortex can be directly transposed to the prototype scale. Indeed, this frequency is only dependent on the operating conditions of the machine. As illustrated in Figure 26, the precession frequency predicted by model tests and the ones identified on the prototype shows perfect agreement. This result is of importance as it demonstrates that the precession frequency of the vortex can be predicted over the complete part load operating range at the prototype scale by measuring the corresponding model at only a limited number of operating points. The values of the first eigenfrequency of the system computed numerically are also compared with the values identified during field tests. Both curves features the same shape, which demonstrates that the influence of the output power at a given value of head is well reproduced by the developed methodology. However, a slight discrepancy is observed between both curves, with a maximum difference equal to 25% of the experimental value.

Power plant network integration
The trend for largescale integration of New Renewable Energy (NRE) sources, such as solar or wind power, characterized by a high degree of variability, creates several challenges in power system operation. NRE variability is a source of imbalance between generation and load in electric power systems since it is not possible to know (with certainty) NRE generation levels for the next hours/days. These imbalances require additional operational reserve to securely handle system operation, thus leading to increasing operational costs in electric power systems. Under such a scenario, using the flexibility provided by energy storage solutions, Pumped Storage Power (PSP) plants play a key role given the volume of energy that they can potentially store. The extreme value of flexibility in electric power systems with a high share of NREs has motivated the industry to seek technological developments in many contexts. Within hydropower, a technological revolution is happening, where the classical fixed speed pump-turbine designs are being replaced by variable speed solutions exploiting advanced energy conversion technologies based on power electronic interfaces either Full Size Frequency Converters (FSFC) or advanced excitation systems in Doubly Fed Induction Machine (DFIM). These technologies provide huge contributions in terms of hydro unit flexibility for changing the operating point with increased efficiency in turbine mode, while making possible the exploitation of variable power operation in a certain band in pump operating mode.
Following the integration of new PSP technologies in electric power systems (together with the integration of NREs), specific studies are required for system operations in order to evaluate system stability with respect to different grid disturbances and different operational scenarios. In these types of studies, dynamic simulation models of different sources connected to the system are used. Such models have been defined for many decades, making it possible to find a well established state of the art domain for most of the available generation technologies. For the specific case of the new PSP technology, the associated dynamic models are not widely available since in many cases the technologies itself are either not in a mature state or are not yet widely exploited in the industry. In particular, it is important to notice there is no widely available models for the new variable speed technologies that are capable of representing the main dynamic response either in turbine and pump operating modes, while being possible to be straightforwardly integrated into existing electric power system simulation packages. Such models need to represent, with acceptable accuracy, the main dynamic response of the electric power plants in different operating conditions. Nevertheless, for system dynamic stability studies, a high degree of complexity is not required, and a common approach is to use reduced order models based on block diagrams with simple transfer functions for representing the dynamics of the main components of the units.

Comparison between detailed and reduced hydropower plant models
Using a multidisciplinary approach among project partners, it was possible to identify and validate appropriate simulation models for variable speed PSP units. The approach consisted of two main steps. First, detailed simulation models of PSP with variable speed generators (operating as a generator and as pump) were developed using the SIMSEN simulation package, where it is possible to include a detailed representation of both the electric and hydraulic parts of the installation. The SIMSEN software allows identifying in detail key control parameters such as the derivation of gate closing and opening laws as well as the structure of the hydraulic governor with respect to key disturbances such as load rejection transients.
At the second stage, typical hydraulic governors such as PID governors as well as the standard IEEE models such as HYGOV2 and IEEEG3 were used as the basis for the derivation of new reduced order models for variable speed units. The electric part of the variable speed generators are based on a classic phasor model for DFIM and FSFC generators. The development of models for pump operation was also addressed and a simplified model based on an equivalent electric circuit of the hydraulic system is exploited in this case. The simulation results obtained from the detailed simulation in SIMSEN were used in order to properly tune and validate the reduced scale order models that are proposed. As an illustrative example, a comparison of the simulation results obtained with the detailed simulation model and the reduced order model for a DFIM based variable speed hydro unit in pump mode are presented in Figure 27 following a sudden load step variation. The obtained results for the validation of the new PSP models relied on existing PSP data – the 400 MW Aldeadávila II pumped storage power plant located in Spain (river Douro) at the border with Portugal. The obtained results are in very good agreement in terms of rotational speed and active power, even if the head and discharge fluctuations related to the downstream surge tank are not reproduced. It shows that, from the power system stability perspectives, good simulation results can be obtained with a reduced simulation model of the hydraulic system for pumping mode provided that the model includes penstock inertia and realistic pump characteristics.

Variable speed PSP contribution for power system dynamics performance and security
In order to evaluate the technical benefits resulting from the integration of variable speed PSP in electric power systems, a test case was created based on the Portuguese electric power transmission system, which currently integrates more than 5 GW of wind power and where hydropower plays a key role. For example, in Portugal, the electricity consumption was fully covered by solar, wind and hydropower in an extraordinary 107 hour run in May 2016, only possible because of the inherent regulation capacity of hydroelectric units. Additionally, in Portugal, the Venda Nova III PSP plant with 736 MW is in its final construction phase and it includes the largest variable speed generators (DFIM type) in Europe. From the study case perspective, future scenarios based on the progressive refurbishment of some of the existing hydropower plants with pump storage capability were considered. This aimed at assessing the resulting technical benefits through additional flexibility and controllability that variable speed technology can provide in comparison to the classical fixed speed solution.
From the power system transient stability perspective, the adoption of a variable speed pump storage power plant (PSPP) over a fixed speed PSP presents notorious benefits because of the increased power system regulation capability, in particular during pump operating mode. In fact, variable speed PSPPs are capable of providing fast frequency control services through the reduction of the consumed power (while operating in pump mode), thus contributing to the so called frequency containment reserve. To do so, a frequency responsive PSP plant model was developed in PSS/E based on the main outcomes provided by the model identification activities previously described. For illustrative purpose, a fault event was considered in the Portuguese power system at Lavos substation (400 kV busbar), leading to the disconnection of about 1,130 MW of non fault ride through compliant wind farms (Figure 28 illustrates the fault location as well as the considered hydropower plants). The obtained simulation results with respect to the system frequency are presented in Figure 29 for different test cases (Case A is the current operating scenario, Cases B, C and D correspond to the progressive integration of the Venda Nova III, Agueira and Alqueva power plants in the system). The obtained results clearly demonstrate that the integration of variable speed PSPs provide a positive benefit with respect to the ability of controlling system frequency following the massive disconnection of wind power as a consequence of the 400 kV transmission grid fault.

Participation of reversible hydropower plants in electric power markets
In addition to the technical benefits resulting from the presence of variable speed hydro units in electric power networks with a high share of NREs, new strategies for the exploitation of this technology in electricity markets were also envisioned. After an extensive survey on the operation and regulation of several European electricity markets, a set of optimization strategies was developed to tackle the participation of hydropower plants equipped with variable speed pump units in the Frequency Restoration Reserve (FRR) market. Additionally, a future scenario was considered in which wind farms participate in both the day-ahead (DA) and the FRR markets in coordination with the hydropower plant. Due to the stochastic behaviour of wind speed and direction, it was necessary to use the flexibility of a pumped storage power (PSP) unit to provide two internal services (i.e. acting as a Virtual Power Plant). By allocating a portion of its available power, the PSP is able to smooth the wind power variability and suppresses eventual forecasting errors, at the cost of losing some participation in the electricity markets.
The bidding strategies are divided into three modules:
- a medium term model which optimally allocates the expected yearly natural water inflows over 52 weeks;
- a short term model which is responsible for defining both DA and FRR market bids, based on forecasted information and the PSP/Wind Farm characteristics;
- an evaluation module that applies the bids to a real market environment and assesses the system revenue.
The envisioned methodologies were applied to real data for the Iberian electricity market (MIBEL), using a real PSP unit located in the north of Portugal as the test case.
Considering the standalone operation, the methodology proved its validity from a technical point of view since, even disregarding the amount of water used in the activation of the FRR, the reservoir level deviation from the expected never surpassed 2.0 hm3. Looking from an economic perspective, a comparison between a pump unit with fixed speed and the variable speed PSP unit revealed an increase in revenue of almost 12% when the bidding strategy is applied, boosting the revenue by € 12.33M. It was also possible to conclude that due to the low liquidity of the FRR market and the considerable capacity of the PSP, the reserve bids were affecting the market clearing price of the FRR market. After applying a tool of market dispatch simulation, developed in this project, it was possible to assess a decrease of 25% in revenue when compared to the expected value. By capping the size of the FRR bids, it was possible to tighten the difference between the expected revenue and the realized.
Due to the stochastic nature of the wind power, a technical validation block was introduced into the evaluation module, which identifies for each instant if all contracted services plus the internal ones were or not fulfilled. Simulation results revealed a success rate (or reliability in reserve provision) of more than 99% in the provision of all services, resulting in only 2 hours, out of a total of 7088, where it was not possible to guarantee the total fulfilment of the electrical energy and FRR bids. From an economic point a view, the coordinated strategy was able to avoid the payment of penalties due to imbalances between contracted and delivered products. However, due to the provision of the internal services, which require the apportionment of some portion of the PSP’s power, the overall system revenue was lower (4%) in comparison with the standalone operation. Additionally, after applying the aforementioned market dispatch tool, the overall system revenue suffered a severe reduction (20%) due to the low liquidity of the FRR market in comparison with the size of the FRR bids. Consequently, under these circumstances, the coordinated strategy becomes more profitable than the individual operation of the PSP and wind farm, although it only corresponds to a 4% increase in revenue.
Potential Impact:
The various impacts of the HYPERBOLE project and their scientific, economic or environmental importance are described in the following chapter. The impacts are distributed for public and private sectors, their presentation being structured as follows:
- Impact on the development of renewable energy in Europe;
- Impact on the European industry;
- Impact on science and technology.
Globally, the HYPERBOLE project decisively contributes to the development and the integration of renewable energy sources in order to achieve the corresponding strategic goals set by the European Union in its agenda 2020.

Impacts on the development of renewable energy in Europe
Both the massive development of New Renewable Energy (NRE) and electricity market deregulation are driving the development of a highly responsive electrical energy system; a network that can match the demand of peak power and take advantage of low prices when the demand is low yielding an excess of electricity. To this end, electrical utility traders require enhanced availability and flexibility of power plants with large-scale energy storage capability. The TSOs need fast transitions in the electrical grid, driving the necessity for power plants to be capable of providing regulation ancillary services in terms of primary and secondary control reserves. Both open and combined cycle gas fired and hydropower plants feature such a capacity for flexible generation at peak power and regulation ancillary services at a large scale. However, hydropower plants also offer the advantage of exploiting renewable primary source energy with almost no emission of greenhouse gas. Additionally, pumped storage power plants provide a system to store large amounts of electricity, with a full cycle of pumping and generation, which can achieve high level of efficiency.
In 2010, 20,225.284 TWh and 3,610.298 TWh of electricity were generated in the World and in Europe, respectively. The geographical breakdown of the worldwide electrical energy generated in 2010 by the type of power plants is shown in Table 1. The total net generation accounts for the total generation deduced from the energy consumption by hydroelectric pumped storage power plants.
An overview of electricity generation distribution by type of power plants for the World and Europe region is reported for the last 3 decades is presented. The key points of the figure are as follows:
i) During the last 3 decades, the electricity produced worldwide has grown by an average annual rate of 3.14 %, while Europe production has increased only by 1.76 %.
ii) Conventional thermal electricity generation represents two-thirds of the total. In Europe, the development of nuclear power during the eighties, leading to a 25 % production share in 2010, has limited the conventional thermal part to less than 50%;
iii) Electricity generation is strongly correlated with the economy as can be seen by the large drop after the 2008 financial crisis.
iv) Hydropower brings a significant contribution; around 17 % share for both the world and Eu-rope, and thus represents the main renewable energy source for electricity generation.
v) The contribution of NREs becomes noticeable during the last decade, especially in Europe where a 10 % share of the total generation was achieved in 2010.
The 20-20-20 strategic energy policy decided by the European Union and its accompanying Renewable Energy Directive requires the EU to generate 20% of electricity from so-called NREs by 2020. With the ratification of the Paris Agreement on climate change (UNFCCC COP21), this energy policy will be reinforced in the future. For instance, a proposal was published by the Commission in November 2016 to ensure a target of a least 27% renewable energies in the final energy consumption in the EU by 2030.
This policy relies on a tremendous capacity increase of both wind and photovoltaic power plants. However, the intermittent nature and uneven localization of NRE power plants can be source of disturbances for the power network. The integration of NREs in the existing power system poses a serious threat to the overall stability of the grid and is therefore slowing down further NRE development. The extension of the operating range of both hydropower plants with generating units and pump-turbines can be exploited to provide frequency regulation capabilities as well as secondary reserves.

The technological novelty of the variable-speed pumped storage facility represents an even more powerful tool in grid regulation due to its high operational controllability and adaptability. As a reminder, the four main features of variable-speed pump-turbines are:
i) Efficient part-load operation in generating mode;
ii) Active power control capabilities in pumping mode;
iii) Improvement of network frequency regulation through improved inertial response (instantaneous active power injection into the grid);
iv) Reduction of oscillating behaviour after a network fault due to high controllability.
The level of NRE integration into the power system can thus be increased steadily with the operating range extension of existing power plant and with the installation of new variable-speed pumped-storage power plants and/or the modernization of existing hydropower plants featuring pump-turbines in on-off operation. The massive development of NREs within the European electrical generation mix will consequently be backed up by both conventional hydropower plants and pumped-storage power plants, either through the improvement of existing equipment or through the development of new types of variable-speed pumped storage power plants. An additional benefit is the mitigation of CO2 emissions generated by combine cycle fired gas power plants.
In addition to the technical aspects of grid stability and security issues of power systems with a high share of NREs, an important economic argument is to be made. As a matter of fact, no suitable business model for modern pumped-storage power plants exists today, taking into account the totality of their provided services over the entire operating range.
In this context, the HYPERBOLE project contributed to the development of an optimization algorithm for the identification of adequate strategies regarding the participation of reversible hydropower plants in the secondary reserve market, in complement to the participation in the tertiary reserve market. Thus, it will be possible for hydro pumping promoters to identify and obtain additional profits resulting from their participation in reserve markets as a result of the extended control flexibility and operation range provided by new hydro units. At the same time, the possibility of exploiting additional new Automatic Generation Control (AGC) signals for fast responding resources such as hydro units, including pump operation mode, was addressed. This new signal allows the exploitation of the full potential from fast responding resources and allows the secondary frequency control to cope with the very short-term variability of NREs. This has not been designed or tested for European countries, and the HYPERBOLE project proposed a new AGC regulation signal, suitable to extract full flexibility from variable speed pump units.
Existing economic studies are mainly focused on evaluating the value of price arbitrage (i.e. the price difference between peak and valley operation) and the value of the provision of tertiary reserve, as it is already provided by the conventional on-off operation of pump-turbines. If the important role of variable-speed pumped storage power plants is to be consolidated and developed, their integration into the pre-sent economic system is to be studied along with their technical integration, in order to ensure a sustainable way towards the increasing use of NREs in Europe’s energetic future on all levels: technical, economic and ecological.
The success of the HYPERBOLE Project will strongly affect the 20-20-20 strategic energy policy decided by the European Union and its accompanying Renewable Energy Directive in leading toward a dramatic electric energy system transition.
The HYPERBOLE project provides the first economic study of how the profitability of variable-speed pumped storage power plants can be assured, taking into account their role in the provision of a secondary reserve to the grid and in primary frequency regulation.

Impact on the European industry
The installation of new capacity and the refurbishment of existing power plants in the course of the development of renewable energy sources and their integration into the grid holds a substantial business potential for the three main hydro equipment suppliers in the HYPERBOLE consortium, namely GE Renewable Energy, ANDRITZ Hydro and VOITH Hydro. They presently hold over 50% of the global market share in hydro equipment, which makes the European industry a key player in the future of hydropower.
The worldwide electricity supply relies primarily on conventional thermal power plants, mainly coal fired; coal alone representing 40 % of the total generation in 2010. In Europe, the development of NRE capacity has superseded the other types of fuels in the last decade as a consequence of public energy policies in several EU27 states. In 2010, almost 25 GW of NRE power plants, mainly wind power plants, have been installed. The peaks of conventional thermal power plants exceeding 10 GW per year in 2002, 2005 and 2008 are strikingly due to new constructions of large gas fired combined cycle power plants.
The key points of the installed electrical capacity data are as follows:
i) Almost 3,000 GW of capacity have been installed in the last 3 decades to create a total of 5,067 GW worldwide in 2010; 437 GW of capacity has been installed for a total of 982 GW for the same time period in Europe.
ii) Conventional thermal power plants are maintaining a 65 % share of the worldwide installed capacity in the last 3 decades;
iii) The reduction of conventional thermal power plants from 65% to less than 50% in Europe was achieved by installing nuclear power plants during the eighties and developing NRE power capacity up to 146 GW in 2010.
iv) In 2010, the worldwide installed capacity of hydropower plants, excluding pumped storage, represents 918 GW, with 168 GW in Europe.
v) In 2010, the hydroelectric pumped storage capacity reached 121 GW worldwide. Europe, with a 45 GW capacity, displays the largest share at 37% of the worldwide pumped storage capacity.
An estimated hydropower capacity of 1,000 GW is to be installed by the year 2050, mainly in Africa, Asia and South America, representing a massive potential contract volume for European hydroelectric machine manufacturers. It goes without saying that in this highly competitive area, an important and continuous R&D effort is necessary to keep an advantage over non-European competitors in the installation and re-furbishing of hydropower plants.
In the modernization of existing hydropower plants, adding just one unit with modern technology could be an economically and technically interesting option. From the presented numbers, it can be estimated that a total capacity of 1,000 GW will need to be refurbished over the next decades, adding to the potential business case for the production of hydroelectric machines by the equipment suppliers participating in the HYPERBOLE consortium.
The combination of the resources and the expertise from all the partners in the HYPERBOLE consortium creates the critical mass needed for such an extensive and complete research project. The obtained results are of upmost importance in the further development of high-end hydro-mechanical equipment. The present configuration of the consortium and the collaboration between all the involved parties offer the European players a unique and unparalleled access to a completely new set of data and insights.
The important knowledge and tools developed in the HYPERBOLE project represent a major competitive advantage for the European hydro equipment industry and allows the partners to defend and extend their share of the global market.
The installation of new and refurbished hydropower plants along with the development of renewable energy sources represents a significant business potential for European manufacturers (est. 1,000 GW to be installed and 1,000 GW to be refurbished by 2050).

Impact on science and technology
As illustrated by the world power per unit record at the Xiangjiaba hydropower plant, the road to the 1 GW turbine has been paved. An absolute control of stability over a large operating range is essential, since the smallest fluctuation in power output can cause critical perturbations in the connected grid, especially with the long transmission lines in many developing countries in Asia, Africa and Latin America.
An important result of the HYPERBOLE project is a detailed knowledge of the dynamic loads experienced by the hydro-electrical equipment during transient operation of a conventional or pumped storage hydropower plant. This knowledge contributes to the development of more resistant and longer living components, as well as to the development of resistant material.
In addition, the results of the different milestones have already been integrated into the SIMSEN software for the multi component modelling of hydropower plant dynamics, which has its origin at EPFL and is continuously being developed with the support of the three main hydro equipment suppliers in the consortium. The result is a powerful tool for a holistic analysis of the interaction of a hydropower plant with the connected power system. The improved SIMSEN software will set a new industry standard as a reference in the field.
In the refurbishment of hydropower plants, mixing conventional (mostly existing) fixed speed units with variable-speed solutions based on DFIM (double fed asynchronous generators) or FSFC (Full Size Frequency Converters) increases the complexity of the hydroelectric system. A tight coupling through hydraulic and electrical systems must be taken into account in the design process. Operating FSFC-based solutions with a simple control objective of constant power delivered to the grid may increase the amplitudes of self-excited torque and flow rate oscillations in part load, so the HYPERBOLE project provided a methodology to assess the consequences of optimized control strategies for FSFC-based solutions. Variable-speed solutions using Doubly Fed Induction Machines (DFIM) or classical layout with Power System Stabilizer must be investigated with regard to torque mitigation. The HYPERBOLE project studied the potential of all these solutions to ease and improve the massive integration of NRE’s into power grids by providing stabilization of the grid. Options to make use of the capabilities of a full converter in one unit to partly compensate undesired behaviour of the other units was considered.
In addition, the HYPERBOLE project contributed to the development of comprehensive electrical models of all relevant types of hydraulic and electrical components to investigate most complex coupling effects, identifying the possible beneficial or detrimental interactions with different types of units.
The large number of numerical and experimental investigations scheduled in the participating universities and research institutions has generated a significant academic output on many levels. The formation of PhD students is a key function of the involved educational institutions. This ensures at the same time a continuous supply of highly trained specialists needed in the industry. Furthermore, the findings are also part of scientific publications, which consolidate the leading position of European research institutions in the fields related to the HYPERBOLE project.
The findings of the HYPERBOLE research project will be at the origin of the development of reliable high-end hydroelectric equipment for variable-speed pumped-storage power plants. The use of various systems stabilizing the interaction with the power system is studied and evaluated. The existing SIMSEN modelling software has been improved to be used as a reference tool for the simulation of hydropower plant dynamics.
The HYPERBOLE project produced an impressive academic output in the form of scientific publications and doctoral theses that ensured the formation of highly trained specialists needed in the industry and the public sector.

List of Websites:
https://hyperbole.epfl.ch

Related information

Reported by

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE
Switzerland
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top