FRADES II Europe s Largest and Most Powerful Doubly-Fed Induction Machine A Step Ahead in Variable Speed Machines

FRADES II Europe s Largest and Most Powerful Doubly-Fed Induction Machine A Step Ahead in Variable Speed Machines Thomas Hildinger Ludger Ködding Philipp Eilebrecht Alexander Kunz Holger Henning Voith Hydro Holding GmbH & Co KG, Heidenheim, Germany Voith Hydro Holding GmbH & Co KG, Heidenheim, Germany Voith Fuji Hydro K.K. Kawasaki, Japan Voith Hydro Holding GmbH & Co KG, Heidenheim, Germany Voith Hydro Holding GmbH & Co KG, Heidenheim, Germany Introduction Pump storage power plants have been part of energy grid systems for more than 100 years. These power plants gained significance in recent years, as they play an important supporting role for the integration of renewable energy sources by providing frequency and voltage stability. In the past few years, the development of frequency converter technology made it possible to apply variable speed generator-motors of large power input/output, which lead to unique advantages for grid stabilization due to the application of this variable speed technology. This development will be exampled at the pump storage power plant project Frades II in the northwest of Portugal, one of the recent projects developed by the main electric generation company in Portugal EDP- Gestão da Produção de Energia. The use of a variable speed motor generator generally offers advantages that can justify the higher investment. These include, among other benefits: Power control in pump operation Extended turbine operating range Fast regulation Dynamic network support The Frades II pump storage power plant has been installed in order to react with the highest flexibility to equalize the volatile generation of electric power from wind farms. By introducing the variable speed technology, the power absorption in pump mode from the grid can be adjusted, and transition quicker, when compared to a classical pump-turbine application. Frades II is equipped with two, 400 MW class pump-turbine units with Doubly-Fed Induction Machine (DFIM) technology for the speed variable motor-generators. These are, from construction point of view, induction motor-generators, but with the rotor power controlled by an AC-excitation system (Voltage Source Inverter (VSI) with DC-link) and transferred by slip rings. As the stator is connected to the grid (via the step-up transformer) and the rotor fed by the AC-excitation converter, the machine is called "doubly-fed." The behavior of such a doubly-fed unit is comparable to conventional synchronous hydro generators, as they are also doubly-fed (e.g., the reactive power output is controlled by the rotor current); however, the rotating speed is controlled by additional regulation of the rotor voltage frequency and amplitude. This paper is intended to overview the large scale 420/433 MVA DFIM variable speed motorgenerators for the Frades II pump storage hydro power plant. Page 1

The construction of a doubly-fed asynchronous motor generator of this size entails its very own challenges. The rotor has been completely redesigned. Instead of fitting the rotor poles to the rotor, as in the case of synchronous motor generators, three-phase bar windings are used for the rotor in Frades II, similar to those normally found in the stator. Furthermore, the rotor consists of silicon plates in order to reduce losses and to improve the efficiency. In contrast to the stationary windings in the stator, the windings on the rotor rotate at about 350 rpm to 380 rpm and are subjected to high centrifugal forces. In the case of load rejection, the speed can reach considerably higher values. In addition, the rotor is loaded with significantly higher voltages and currents compared to conventional synchronous machines through a highperformance frequency converter. The converter is 25 times more powerful than comparable excitation systems for conventional fixed speed machines and consequently also larger and heavier. All of this has a decisive effect on the design of the rotor. Therefore, a completely new product was developed for this project. As part of the commissioning phase, Unit 2 of the Frades II power plant's motor-generator' efficiency and electric characteristics were assessed. 1 Mechanical Design Features As mentioned above, although the stator of a synchronous generator and the stator of an induction machine are similar in their design, the main differences are in the rotor construction. As the rotor of a salient pole synchronous hydro-generator is well known, some constructive details of the Frades II DFIM rotor are highlighted in the following subsections. 1.1 Frades II Rotor Design Goals At a cursory glance to a motor generator cross section drawing, a modern DFIM rotor looks simpler when compared to a salient pole rotor. The rotor rim with inserted winding forms one single body instead of the clear separation of the rim and poles of a salient pole machine. However, one of the difficulties in the DFIM rotor design is the rotor winding overhang, with its high voltage insulation and its support against very high centrifugal forces. The rotor winding overhang also results in the bigger axial length of the rotor, which can have negative influence on shaft stability and overall unit dimensions. This feature leads to the first goal of Frades II rotor design: a reliable rotor winding overhang support system optimized for keeping the rotor as compact as possible. The second goal was, as the cross section suggests, to develop a simple rotor structure that is safe and efficient to erect while allowing good access for maintenance, specifically easy access to rotor bars. Page 2

Fig. 1 Frades II Rotor Assembly Rotor winding overhang support Rotor core Rotor core clamping bolts Radial ribs Head beams Clamping plates Brake ring These two goals had to be realized in conjunction with the obvious goals and boundaries of the project, such as combining reliable mechanical operation with excellent electromagnetic and ventilation performance. 1.2 Rotor Hub The Frades II rotor hub (see Fig. 1) is a radial rib type fabrication constructed using welded, lowcarbon steel plate. Its design is similar to a synchronous machine. Longitudinal keyways (head beams) made from U-profiles are welded to the outer side of the ribs and accept keys to transmit the torque between the rotor hub and the rotor yoke. They are designed to also take the forces occurring on short circuits. The keys ensure the rotor yoke remains in a concentric position at any speed. Hooks on the bottom of head beams carry the weight of rotor core/yoke and the rotor winding. Additional to the rotor core, the rotor winding overhang fixation system is also guided by the head beams, requiring them to be extended axially, resulting in a significantly longer rotor hub than for a synchronous machine. These upper and lower portions of the radial ribs, together with the torque transferring reinforcement disc segments between them, are specifically designed to allow the cooling air to flow towards the rotor core ventilation ducts. Removable covers further help guide the air inside the rotor hub while allowing access for maintenance. On the lower side, the brake ring forms the outer diameter of the air inlet into the rotor hub. Its fixation is integrated in the rotor hub. Page 3

1.3 Rotor Core As in a synchronous machine, the rotor core has to withstand the centrifugal forces and support the rotor winding. It additionally acts as a flux ring. The magnetic field rotates relatively to the rotor, requiring a silicon steel quality with low magnetization losses in thin sheets to reduce eddy currents. Typical silicon steel sheet qualities used in the stator core do not have guaranteed mechanical properties. For the Frades II rotor core, a close cooperation with the steel supplier allowed us to create and specify a rotor core material fulfilling both the electromagnetic and the mechanical requirements. This successful integration of the two functions as a magnetic flux ring and as a structural element greatly simplified the overall rotor design. A mechanically solid body has to be formed from the thin steel sheet segments, with ventilation slots to allow for the cooling air to pass and cool the rotor winding. The spacers keeping the ventilation slots open have been fixed mechanically against the centrifugal forces, safely securing them from entering the air gap. The rotor core is compressed by numerous high tensile clamping bolts, with non-magnetic endplates on both ends of the rotor core evenly distributing the clamping force to avoid a squirrel cage effect, clamping bolts are insulated against the core and end plates. The bolts are designed to keep the required pretension for the entire lifetime of the unit. The bolts are accessible for checking and could even be retightened without dismantling the rotor. On the inner diameter of the rotor core, tangential keys form the connection to the rotor hub. These keys transfer torque, support against unilateral forces, and keep the rotor core concentric to the shaft while allowing for radial expansion due to the temperature and centrifugal forces. Being loaded by centrifugal force, the rotor core of a DFIM has the same requirements for fatigue life design as a conventional salient pole machine. Extensive finite element calculations have been performed for optimization of the mechanical design. In addition, fatigue tests were done on a model of the stacked body, confirming the assumptions made during the design. 1.4 Rotor Winding The Frades II DFIM rotor is equipped with a three-phase high voltage AC winding that is comparable to a high voltage stator winding. The rotor winding is composed of 360 transposed Roebel bars with two layers per slot representing the maximum safety against short-circuit between the turns. Each turn is formed by separated bars with the effect that rapid voltage changes as generated by the AC-excitation system are acting on the bar s main insulation only. The rotor main insulation made of proven Micalastic insulation system in VPI technology is perfectly able to withstand the specific loads of a DFIM rotor winding (e.g., the voltage increase during transient operation as well as special voltage waveform and peaks due to the switched converter feeding). The rotor winding is insulated with materials to fulfill the thermal class 155 (F) according to IEC 60034.1. For the fixation of the bars in the rotor slots, the same proven system as the stator windings is used, with putty wrapped within a conductive paper around the bars. The putty adapts nicely to the air ducts and to the surface, despite the unavoidable stacking irregularities of the rotor slots. After curing, the putty locks the bars tightly into their position, thus preventing the bars safely from any axial movement and avoiding the need for additional mechanical locks. A cross section through a rotor slot is shown in Fig. 2 (on the right side). The individual bars are inserted in the slots, including sliding and filler strips and a semi-conductive layer. For comparison, a typical stator slot is shown on the left side. Page 4

Fig. 2 - Typical Cross Section of Stator and Rotor Slot While a general similarity between the rotor and stator slots cross sections exists, there are some differences essentially caused by the low frequency in the rotor winding. This low frequency allows the use of single conductors of increased thickness compared with a stator bar design. Generally, the rotor has a deeper slot and smaller slot width, hence improving the cooling of the rotor bar. The slot wedge is designed to keep the radial load from the centrifugal force, resulting in a much thicker wedge compared to the stator slot wedge. In the overhang region, this depth, comprising of the slot wedge thickness and the machine s air gap, helps to reach all dielectric clearances required and to gain the space for the overhang bar fixation. As the rotor winding is exposed to high radial centrifugal forces (different from the stator winding) no ripple springs will be used as slot top fillers. Instead of the ripple springs, solid top fillers are fitted in the slots in order to ensure a safe fixation and uniform pressure on the bars under all centrifugal and magnetic forces. 1.5 Rotor Winding Overhang Support System Inside the slot region, the rotor bars are secured towards the radial forces by the slot wedge. In the end winding region (winding overhang), the bars must be kept in place by an appropriate retaining system. To keep the winding overhang as short as possible, the distance between adjacent bars can be reduced to the minimum distance as required for insulation reasons. Being a normal design practice for the stator winding this is only possible for the rotor winding if no retaining elements have to be passed between the bars in this inclined region. Page 5

The Frades II rotor winding overhang support system only passes two rows of bolts through the rotor winding overhang, one right below and one right above the inclined region. Compared to systems with multiple retaining elements passing through the overhang, this helps to reduce overall rotor length and weight and also reduces the rotor winding resistance. Further, it results in an open design with the best possible conditions for cooling. Between the two rows of retaining elements, the rotor bars are supporting themselves against centrifugal and other forces. Extensive effort was made in enabling the overhang portion of the rotor bars to form a mechanically rigid element with a long lifetime under mechanical and electrical load. Tests of model rotor bars (Fig. 3) and of the actual production rotor bars have proven strength and lifetime of the Frades II rotor winding. Similar in approach as for the rotor core, this design enables an electrical component to use its mechanical strength as well, resulting in a simplified design with superior properties and comparably low number of parts and good access for maintenance. Fig. 3: Exemplary Rotor Bar Fatigue Test The retaining elements supporting the rotor winding overhang against centrifugal and other forces are composed of high tensile bolts with low magnetism and of glass fiber plates that transfer the loads from the rotor bars via the bolts to two support rings located on the radial inner side of the rotor winding overhang (Fig. 4). Extensive testing has been performed for these parts as well, including tests for finding the optimal layup, resin, and reinforcement fibers for the glass fiber plates by manufacturing numerous test pieces and performing static and fatigue testing in different environmental conditions. Page 6

Rotor winding support plates Non-magnetic bolts Support ring Fan blade Spring element Fig. 4-3D Model of the Rotor Winding Overhang Support The centrifugal and other forces on the rotor winding overhang are supported by forged support rings that form a comparably lightweight overhang support system. This light weight allows the rotor winding bars to move the entire overhang support system axially when they expand, reaching operating temperature. The forged support rings are guided in the head beams of the rotor hub and can slide axially. To support this sliding and to compensate the weight of the support system, spring packages are arranged below and push the whole system upwards. This allows free axial expansion of the rotor bars, avoiding additional stresses and deformation due to the natural thermal expansion of the rotor winding. Special attention is paid to the shape and its minimized tolerances of the bars in the winding overhang area to ensure its proper and equal support. An additional feature is the open design of the structure that allows cooling air to freely pass through to cool the winding overhang. Fan blades are located between the support rings to support the cooling function. To simulate the site assembly and to train the site personnel, a mockup of the rotor winding overhangs and their support system has been built and assembled in the Voith Hydro workshop (Fig. 5). The assembly steps from bar insertion, lashing of the winding, and arrangement of the support plates to the insertion and tightening of the non-magnetic bolts could be simulated and trained to prepare for a smooth assembly at Frades II site. Page 7

Fig. 5: Rotor Winding Overhang Assembly Test Model 1.6 Mechanical Design Summary By enabling and by full utilization of both mechanical and electromagnetic performance of the rotor core and rotor winding, the two goals of a compact and a simple rotor design could be achieved. This achievement also came in conjunction with a comparably lightweight and rigid overall rotor design, reducing the required crane capacity and improving the shaft system stability. Furthermore, the open design offers excellent accessibility for installation, inspection, and maintenance. 1.7 Slip Ring and Brush-Gear System The design of the slip rings and brush-gear assembly follows, as far as possible, the same principles as a typical Voith Hydro synchronous generator: a) Self-ventilated slip rings: The slip rings are rotating freely in air. Each slip ring is fixed by insulated rods, neither mounted to any kind of hub nor covered. Each slip ring is in full contact with the ambient air (Fig. 6) except at the insulated studs. The rotation of the slip rings themselves creates air movement. The design of the slip rings accounts for the heat transfer from the rings to the surrounding air. Page 8

Fig. 6 Slip Ring Assembly at Manufacturer s Workshop b) Materials similar to standard hydro units: The slip rings are made from steel; the brushes are of low friction type natural graphite. This is the same combination as in many other hydro-generators and motor-generators. The material has been tested and validated for the special requirements at Frades II, such as AC instead of DC, starting sequence, ambient conditions, and current distribution. c) Open ventilation concept air exchange with the ambient conditions: In general, but even more for high power brush gear assemblies like in Frades II, the ambient conditions play an important role regarding the wear of brushes. The brushes and slip ring system shall be operated in an appropriate temperature and humidity range. Closed ventilation loops, incorporating heat-exchangers as coolers, can lead to a rather dry ambient condition, which is not favorable for the patina built on the slip rings. Therefore it was decided to design the system as open concept, with the cold air supplied from the ambient cavern ensuring the temperature and humidity in the correct range. The differences to standard fixed speed synchronous generators are basically in the pure rating (Table 1). FRADES II Comparable typical SYM Number of sliprings 6 (2 per phase + 1 neutral) 2 Rating 7 100 A (AC), 4 000 V 2 100 A (DC), 300 V Number of brushes Up to 288 40 Losses to be removed ~ 60 kw ~ 5 kw Table 1 FRADES II Slip Ring Main Data Whereas 5 kw of losses can be removed from a large slip ring compartment just by natural exchange, this cannot be done with the 60 kw of losses in the Frades II slip ring compartment. Page 9

As mentioned above, the slip rings/brushes themselves are self-ventilating, but the heat must be actively extracted from the compartment. This is realized with a combined heat and brush dust extraction system as shown in Fig. 7. The warm air is taken, together with the brush dust, directly where the losses main part occurs: directly around the brushes and brush holders. The polluted and warm air is collected in two manifold pipes and directed to an external filter unit where the brush dust is separated. The clean air is released into the common powerhouse exhaust system. Fig. 7 Extraction of Heat and Brush Dust The system has proven to reliably collect the brush dust, as well as keeping the slip ring temperature at perfectly low temperature. The brushes wear is in the same range as for typical hydro motor-generators. 1.8 Ventilation The approach for the design of the ventilation was two staged, with analytical calculations for the pre-design and CFD for fine tuning the final design. During the pre-design, some margins in air flow as well as adjusting capabilities were considered. The calculation method that was derived from salient pole machines had to be modified in order to respect the different rotor shape. Special attention was given to the cooling of the overhangs of rotor and stator. A radial fan was added inside the support. The cooling of the rotor overhang created a second flow path that is parallel to the rim ventilation of the stacked rotor core. Such an approach is unusual because two ventilation paths with two different blowers (rim ventilation and radial fans) are not easy to control and can have unwanted interaction. On the other hand, as a DFIM does not have salient poles, the air gap is very small and constant, which reduces the area where the two ventilation systems can influence each other. The rim ventilation is adjustable at the rotor hub inlet similar to the concept for large synchronous hydro-generators. The overhang ventilation can be adjusted at the inlet into the fan, which is integrated in the winding overhang support; see Fig. 1 and 4. A second adjustment is possible at the openings in the stator frame. This way any unwanted air exchange between the stacked part and the overhangs can be prevented. When the mechanical design was at a stage in which the main components were modeled, a full 3-D CFD was executed. Overhangs were modelled, including the rotor and stator's jumpers. The stacked rotor core was modelled completely too; no equivalent porous media were used. This Page 10

was done to verify, for instance, the impact of the special design of the ventilation ducts on the rotor core. The much higher axial clamping pressure in the rotor core, as compared to a stator core, required stronger duct spacers that also had to be secured mechanically to avoid loosening even under the most severe conditions. After the initial verification, several iterations were run with seemingly small adjustments that did help in optimizing the ventilation system. For example, for the overhang air flow, the design was adjusted to a minimum exchange between the iron part and overhang. 2 Frades II Prototype Experiences 2.1 Erection at Site Frades II is a cavern power house with the typical transport limitations for this kind of power station. The rotor hubs could be transported in one piece; however, the stator frame had to be segmented. Two identical power units are arranged in the power house, composed of doubly-fed variable speed motor-generators connected to reversible pump turbines. The main transformers are arranged in an elongated part of the main cavern, the AC excitation and their transformers are inside tunnels that connect to the draft tube gate side cavern. The erection bay in the cavern is sufficient for stacking/winding two generator rotors in parallel. The stators have been stacked and wound in their respective pits; the provisions for the turbine side assembly allowed the independent erection of the pump turbines. Stacking of the rotors was done with a mobile stacking platform moving around the rotors on a rail track (Fig. 8). Continuous stacking with high material deposit rates was achieved in a very safe and controlled process. Fig. 8: Rotor Stacking Platform Rotor winding bar insertion was performed using a special fixture for handling of the very long and thin rotor bars, avoiding deformation or other damage. The Frades II rotor winding overhang support system is designed to be installed step-by-step, with installation of support blocks and spacers after the insertion of the bottom bars, and final support plate installation after top bars insertion. The lashing of the rotor bars is also applied step-by-step, with good accessibility of the bottom bars overhang from radial inside facilitating these works. Page 11

Fig. 9: Rotor Winding Overhang Support During Site Installation By parallel stacking and winding the two rotors, installation of the second rotor was achieved in less than three months after the first one. The very small air gap required tight roundness and dimensional control during stacking and winding, and very high precision and perfect preparation of the rotor insertion manoeuver. Fig. 10: Frades II Rotor Insertion 2.2 Prototype Testing For large hydro-generators and motor-generators, in general, a well-known procedure has been established to measure the relevant machine data, such as performance and reactances. These procedures have been developed and proven, and take the special requirements in hydro business into account, such as site assembly of the large units and the direct coupling to the hydro turbine. They are described in several national and international standards, such as IEC60034-2. Page 12

However, these procedures do not take into account special requirements and particularities of speed variable units, in particular of a DFIM like Frades II. Such motor-generators are doublyfed; same as standard synchronous hydro generators but the machine is constructed as an induction type. It behaves like a synchronous machine, just at different synchronous speeds, however it looks like an induction machine. With its two, three-phase high-voltage windings magnetically coupled by the stator and rotor iron core, it may to some extent even be seen as a transformer. On the other hand, typical test procedures developed for induction motors do not apply as they do not consider the site requirements. Most induction motors are type tested. Even in case of prototype measurements, testing is typically performed in the workshop on a test bench, as such motors are much smaller than hydro machines. Therefore, a test approach has been developed that combines the strength of induction motor and synchronous generator testing, and applies a measurement procedure that is applicable to site conditions. The basic approach for all measurements was to keep them as close as possible to well-known practices and to the proven testing for standard synchronous hydro-generators. In the following subsections, some of the measurements performed at Frades II are described. 2.2.1 Locked Rotor Test Locked Rotor Tests (LRT) are well-known, described procedures for testing induction motors (e.g., in IEEE Std 112). The basic idea behind that test is that the behavior of a locked rotor induction machine is similar to a transformer. For transformers, the parameters of the equivalent circuit diagram are commonly determined by short circuit and open circuit tests. In these two tests, one side of the transformer is fed by a voltage source, and the secondary side is in shortcircuit, respectively in open circuit. By measuring the currents and voltages, the electromagnetic main parameters of the transformer can be determined. Doing the same with a large induction machine like Frades II has some challenges, as neither the stator nor the rotor could be connected to the grid at standstill condition without causing major complications. For the Frades II, an LRT procedure has been developed that takes the following features into account: a) Failure safe rotor locking with the installed machine brakes: Motor-generators such as Frades II are often equipped with a brake system that is sometimes combined with the rotor lifting system. This system is typically designed to keep the power unit at standstill even when some water leakage through the closed turbine wicket gates is producing residual driving torque, typically in the range of a very few percent of the rated torque. During the LRT, the feeding parameters (e.g., voltage and current, and the frequency) must be carefully determined and observed as the unit can easily provide a driving torque much higher than any brake could hold. For safety reasons, it was decided to test with reduced parameters in order to not exceed the installed braking torque. In case of a failure, no major problem to occur; some short slipping of the installed braking system would not cause any major damage. b) Feeding the rotor with the AC-excitation system: As mentioned above, there is no option to connect the machine to the grid during the LRT. Therefore, the AC-excitation system, basically a three-phase voltage source converter, was used to feed the machine. Different from the standard LRT, the converter is connected to the Page 13

rotor. This approach makes the test setup easy and safe, but the stator voltage does not reach very high values for typical configuration and frequency. For Frades II, it was limited to approx. 0.1 pu stator voltage, which results in lower accuracy of the test and furthermore it must be accepted that the reactance, the main field reactance, cannot be obtained in saturated mode by that approach. c) Stator winding in open- and in short-circuit mode: Both tests have been performed to the Frades II motor-generator using the installed shortcircuit switch. This is a breaker with a short-circuit bridge, which is in operation during electrical starting and stopping sequence of the unit. d) Different stator-rotor positions tested at standstill: The position of the rotor relative to the stator can influence the measurement results. The IEEE standard recommends a certain procedure to find the optimum (average) position by slow movement of the rotor prior to the LRT. This movement is difficult to realize for large machines like Frades II, so the LRT is performed at different positions instead. Under the given test environment and accuracy, the Frades II results did not show some reproducible dependency of the rotor position, but in general it cannot be excluded. The macroscopic behavior of the DFIM can be simulated with help of an Equivalent Circuit Diagram (ECD) as shown in Fig. 11. Fig. 11 ECD of the DFIM For the LRT, the machines slip s equals 1. The converter is connected to the secondary side feeding the voltage U 2. By measuring the voltages and currents at a known frequency, the first values for the transformer ratio (ueh) and the sum of the longitudinal reactances can be obtained by the shortcircuit test. An open loop test determined the current I 0, resulting in the value for X h. The resistor R Fe, representing the iron losses, cannot be determined separately by this test, as there is no practical way of separating the sum of the longitudinal reactances in stator (X 1σ ) and rotor leakage (X 2σ ). The Frades II LRT was performed with 12.5 Hz frequency. Lower frequencies can, during the closed loop LRT, help to reach higher rotor currents with the given equipment, but can also quickly lead to high driving torque. As the brake capacity (rotor locking capacity) in Frades II was limiting the torque, it was tested up to approx. 4 ka rotor current, resulting in approx. 8.5 ka stator current at 2 kv rotor voltage hence not reaching the rated data. A second test with 50 Hz allowed higher rotor voltages of about 4 kv (approximately the rated rotor voltage), but with lower currents during that test. Even with both tests not reaching rated data, the obtained results are similar. Page 14

Fig. 12 LRT Short-Circuit Test Parameter at 12.5 Hz and 50 Hz The LRT was performed in the very beginning of the commissioning phase with the intention of obtaining the first feedback of the pre-calculated machine data. This is the first time the converter has been connected to the machine, supplying high currents; hence, the test is valuable as part of the commissioning process (e.g., to check the brush-gear system, the stator and rotor data acquisition, etc.). The LRT can be performed during dry commissioning phase without the turbine, and thus without water available (except the cooling water supply). As mentioned above, there is some risk of the machine torque reaching high values during the test. This may happen if the pre-calculation leads to a wrong assumption, the converter has a malfunction, or an operational failure occurs during the test (converter in manual operation), which is even more likely considering the very early phase of commissioning, with some protection and automatization systems not yet fully available. Therefor, provisions must be made so that that starting rotation would not cause any harm. The bearings should be ready for rotation, any movement should be observed. It must be possible to stop the test immediately in case of trouble at any location along the whole shaft-line, including the brushes and slip rings assembly. This requires an arrangement of a preliminary emergency stop system if the final system is not yet in place. In fact, the LRT at Frades II was not intended to obtain the machine parameters with the highest accuracy, as the whole measurement setup was trimmed to get a fast overview, to verify the calculated data for the installation and commissioning procedures, and to check the converter and motor-generator for the first time under current loading. It was not possible to reach rated values for currents and voltages during all LRTs due to the given restrictions, site conditions, equipment, and machine parameters. Following these intentions, the LRT at Frades II delivered excellent results and the setup was proven to be an easy and practical way to get a quick verification of the machine and its ECD parameters. Although it is not possible to follow the mentioned IEEE standard for LRT in all details, the test was valuable and the results were close to the final measurements. The highly accurate determination of the machine data was performed later during the prototype testing, when the machine was readily commissioned (see subsections 2.2.2 and 2.2.3). Page 15

2.2.2 Machine Parameters As mentioned in subsection 2.2.1, the machine behavior during normal operation, as well as in failure modes such as short-circuits, can be simulated by means of the ECD, as shown in Fig. 11. This means, same as standard synchronous units that are described by the well-known reactances (xd, xq, x d, x d, ), the DFIM is described by the ECD, respectively by the elements of the diagram. Therefor it is important to accurately and reliably pre-calculate that element data, depending on the machine design, and also to verify the values after the installation of the machine. In general, the available standards for machines like the large hydro speed-variable DFIM for Frades II are not well established because these machines are quite special within the family of electric machinery. Standardized procedures are available for induction motors, and testing of large hydro machines is well established, but the combination of both as in Frades II is not fully covered by the available standards. Many DFIM are installed in the wind power industry, but their units are typically assembled and tested in the workshop, whereas the large hydro units are assembled at site. For Frades II, the determination of the equivalent circuit data has been developed at Voith based on the measurements of different load-points under open-loop and short-circuit operation. Different from the Locked Rotor approach (subsection 2.2.1), measurements and evaluation have been performed to obtain the ECD data with high accuracy. Therefor, the stator and rotor currents and voltages, including other data like speed, frequencies, and temperatures, have been recorded. The ECD parameters have been recalculated by means of an iterative approach (see Fig. 13) from that measured data. The resistor R Fe represents the hysteresis and eddy current losses in the magnetic material ( iron losses). But, as this representation is neither physically exact, nor useful for different loadpoints, nor really relevant for the typical load-point calculations, it may be omitted in order to simplify the calculations. Beside the resistance R Fe representing the magnetization losses, the parameters of the ECD elements are both the winding resistances R 1 and R 2b, both the winding leakage reactances X s1 and X s2b, the main field reactance X h, and the winding ratio stator/rotor winding called ueh in this report. The relevant tests are the Open- and Short-Circuit Curves (OCC and SCC) and the determination of the no-load losses (for R Fe ). The ohmic resistances R 1 and R 2 are determined directly by measurement. The LRT can give some information, too, but the test was intended to get a first and rough verification of the relevant parameters at an early stage of the commissioning process. The results of the LRT performed approved the calculated data and are similar to the final test results, but the LRT test accuracy is not as exact as the final OCC and SCC testing. A voltage drop at the slip rings is considered between the actual rotor winding voltage and the rotor voltage as measured at the excitation system. Page 16

Fig. 13 Flow Chart: Determination of ECD Parameters The recalculation method developed for this examination is based on an iterative approach, as per the following principle: 1. Calculation of the initial values of all ECD parameters, as per the final machine design. The calculated data are the starting values and are marked with the index 0, e.g., ueh(0), X h (0). 2. The DC winding resistances are replaced by their measured values. The values of the resistances are kept constant during all iterations; however, R 2b (i) = ueh(i)^2 * R 2 changes with the relation factor ueh(i). The measured resistance values are adapted to 75 C temperature, as the machine was warmed up before the SC and OP tests. 3. The resistance R Fe is determined corresponding to the no-load losses. The influence of R Fe of the complete re-calculation is very small. 4. Based on measured data of the SC test, the appropriate ueh is determined. 5. The leakage reactances X s1 and X s2b are determined from the SC test. 6. The main field reactance X h is determined out of the OC test results. The OC curve is used to re-calculate the saturation curve of X h and distinguish between the saturated and nonsaturated X h. 7. The results of iteration number i are the input data for the next iteration i+1. 8. After a few iterations, the differences between the values X(i+1) X(i) reach minuscule values and the process can be terminated. Page 17

The transformation ratio ueh is incorporated to transform the rotor elements of the equivalent circuit diagram ECD to the stator side. The value of ueh (sometimes referred to as NN ssssssss NNrrrrrr or NN 1 NN2 ) can be exactly calculated by the arrangement of the stator winding and the rotor winding. These data can be obtained from the relevant winding diagrams. Therefor, the design value of the ratio is defined and would not need to be measured. Nevertheless, in this calculation the ratio is obtained from the measurements. This measured value is close to the design value, but not exactly the same. As the design value is supposed to be exactly the value of the realized machine, the difference between both can be interpreted as non-accuracy of the measurement as well as non-accuracy of the equivalent circuit diagram itself. It is well known that the ECD does represent the machine only to some extent. Some effects are not considered in the ECD, such as AC-resistance of windings, capacitive couplings, additional losses, and saturation effects. For most calculations, such as load-point assessment, short-circuit behavior, or failure torques calculations, even under non-symmetrical conditions, the ECD offers sufficient accuracy. For other investigations, such as high frequency response or higher harmonic analysis, the simple ECD as described in this paper is not valid. Hence, recalculation of the ECD parameters out of measurements, even under theoretically perfect conditions without measurement uncertainty, cannot result in error-free parameters, as the ECD itself is only a simplified picture of the machine. This may, for example, lead to a non-converging iteration process or other deviation from the calculated/measured results to the realized parameters. The investigations of the Frades II test results and re-calculation of the ECD parameters showed a smooth convergence both overall and for the segregated parameters. Therefor, it is assumed the results are accurate and reliable. The separation of the total leakage X s reactance into stator and rotor leakage can be done in different ways; nevertheless, this separation is not directly based on the measurement. A simple approach would be to split the total value into two equal halves. This is a regular approach and typically done for transformer analysis. As the stator leakage on one side, and the stator related rotor leakage on the other side, are typically quite similar, the simple approach is reasonable; nevertheless, the split for Frades II has been done based on the calculated design values. Page 18

Fig. 14 Iteration Results of the Re-Calculated ECD Parameters, Compared with the Design Values Fig. 14 shows the parameters as per each iteration step. As shown, no significant changes can be observed after approximately three iterations and the parameter values stay stable. In general, the deviation from the pre-calculated design data are very small. The saturated main field reactance is deviating from the design value by approx. 14%, which is still a good result. All other ECD data, including the ohmic resistances R 1 and R 2, are perfectly matching (deviation below 7%). 2.2.3 Efficiency Test The efficiency test has been performed based on the water-calorimetric method by measurement of the segregated losses as per IEC60034-2, which is the same as typical synchronous machine hydro-generators. Each run for ventilation, open- and short-circuit losses was executed at three different rotational speeds. The speeds for the ventilation measurements were spread over the whole speed range while the open and short circuit losses were done under-synchronous and below the dead band of the rotor converter. This was done based on the fact that the losses develop similarly with the difference of speeds in under- and over-synchronous mode. The tests were performed in Generator-Mode, which is usual for pump storages since it is not possible to measure segregated losses in Motor-Mode. Two, three channel power meters were used to record electrical values from rotor and stator simultaneously. For the rotor voltage measurement, voltage dividers were used and Rogowski- Coils for the current as there are no classic voltage and current transducers as found on the stator. The power meter has to have the capability to be used with converters. The method of adding the segregated no-load, short-circuit, and ventilation/bearing losses does not exactly cope with the DFIM arrangement, as some additional losses due to the AC feeding of the rotor current are measured twice in the no-load and short-circuit test. They are partly included in both tests, but cannot easily be separated nor extracted. Theoretically, an approach to overcome this may be to separate the rotor current into active and reactive loads, but this is difficult to do due to the clocked voltage signal of the converter, and thus it was not followed at Frades II. Therefor, the sum of measured losses is expected to be a little too high. The Page 19

difference can be calculated, but then the results would not be a pure measurement-based result anymore. For Frades II it turned out that, even with some losses included twice, all efficiency guarantees were fulfilled. In other words, either the losses that were included twice are negligible, or they are compensated by others. Fig. 15 Simplified Power Flowchart in Generator Sub-Synchronous Operation The calorimetric measurement runs have been performed three times each in order to get information about the dependency of the losses on the speed. The losses of the excitation system are included in the efficiency, which is the same as for standard synchronous generators. These excitation system losses comprise of the losses in the converter, including crowbar, which is the largest part, and also the losses of all other elements in the excitation circuit, such as excitation transformer, in the cables and bus ducts, current limiters, and the slip ring system. Not all of these components losses can be directly measured. The converter losses are determined in the workshop and adapted to the relevant load-points. Different from typical hydro units, the excitation system losses for the DFIM Frades II arrangement equal approx. ¼ of the total losses. The losses of the excitation system must be determined based on the relevant loadpoints at the low voltage side of the main transformer. In case of reactive power supply from the grid-side converter this produces additional losses, which have to be properly considered. Fig. 16 Typical Components of the DFIM Excitation System Circuit Page 20

The excitation current (rotor current) of the DFIM has been calculated by the ECD and the measured ECD parameters as per subsection 2.2.2. Fig. 17 shows the vector diagram that determines the relevant rotor data. Fig. 17 Vector Diagram ECD Currents and Voltages The rated efficiency measured at the DFIM Frades II significantly exceed the level of 98.4%, including the losses of the both generator guide bearings and including the losses of the ACexcitation system (Fig. 16), which is a remarkably high value for a speed variable DFIM application and well above the guaranteed limit. 2.3 2000 Hours Inspection A thorough inspection was performed following 2000 hours of successful commercial operation of the Frades II units. Main goal of the inspection was to check the condition of the rotor, especially of the winding overhang and its support. The accessibility of the rotor winding overhang after disassembly of some covers allowed easy inspection of the rotor winding overhang partly from the outside, from the top/bottom, and also from the radial inside (Fig. 18). Page 21

Fig. 18: 2000 Hour Inspection: Rotor Winding Overhang Radial Inside, Spring Package The rotor winding overhang was found to be in excellent condition, with no signs of looseness, movement or deformation. Also no discoloration or other signs of high temperature were found. The winding overhang support, including the spring package system (Fig. 18), were also in as-designed condition. All gaps and part positions were correct, indicating correct function of the system. Special attention during this inspection was also paid to the compactness of the rotor core, as a certain setting of the almost 4 m long stacked body from thin laminations cannot be ruled out even after applying the specific pre-setting method developed for Frades II. For verifying compactness, a statistically relevant number of rotor core clamping bolts were untightened and checked for remaining pretension. The result was excellent, with almost the complete design pretension remaining and with the untightened clamping bolts still being movable by hand in the tightly tolerated holes of the rotor core. This clearly indicated that no relative movement had occurred within the rotor core even during the full transient overspeeds reached during the commissioning tests, with the laminated rotor core behaving like a single solid cylinder. It also proved the precision of the stacking and erection works; a remarkable achievement by the field service team. Of special interest for the power station operator was the condition of the slip ring and brushgear system. The first impression when entering the slip ring room was the cleanliness: no accumulations of brush dust could be found, indicating that the dust extraction system is working as designed. The wear of the brushes is within normal range and was uniform among all brushes. The surface condition of slip rings and brushes was good. Fig. 19: 2000 Hour Inspection: Condition of Slip Rings and Brush-Gear Page 22

The Frades II variable speed motor generators have presented themselves in excellent condition at the 2000 hours inspection and could immediately go back into service. The operation logs show that the capabilities of the variable speed units are extensively used by the operator, with the speed being varied frequently. 3 Conclusions Equipped with two pump-turbine units of the 400 MW class and the state-of-the-art technology for the speed variable the motor-generators rated 420/433 MVA, the DFIM machines of Frades II are Europe s largest and most powerful of its type. Comprehensive investigations and tests have been performed during the conception phase as well as the design phase in order to ensure a safe, robust, and reliable state-of-the-art machines that meet a wide range of targets, from easy and simple access and maintenance to high efficiency, and the highest operational flexibility with a fast, dynamic response to the grid's needs. The tests methodology for these machines had to be adapted for this application as the existing test methodology for both classical synchronous machines typically used in hydropower generation as well as for large asynchronous machines used for industrial application have not been developed for the DFIM as applied in Frades II. The measurement showed excellent results for all measured parameters, including reactances, losses/efficiency, and temperatures that were in line with the predicted figures and fulfil the guaranties. Last but not least, the Frades II motor-generators have presented themselves in excellent conditions and went immediately back in service after the 2000 hours inspection. The operation logs show that the capabilities of the variable speed units are extensively used by the operator with the speed being varied frequently. 4 References Hildinger T., Ködding L., Modern Design for Variable Speed Motor-Generators - Asynchronous (DFIM) and Synchronous (SMFI) Electric Machinery Options for Pumped Storage Power Plants, Hydro 2013, Innsbruck, Austria Koutnik J., Hildinger T., Bruns M., Ein Sprung nach vorn drehzahlvariables Pumpspeicherkraftwerk Frades II, WasserWirtschaft 5 2015 Schwery A., Sari G., Guillaume R., Kunz T., Experience From a Large European Pump Storage Plant in Operation with Variable Speed Units and Areas Where the US Grid Can Benefit From the Knowledge Gained, Hydrovision 2017, Colorado, USA IEC 60034-1 Rotating electrical machines Part 1: Rating and Performance Edition 13 2017.05 IEC 60034-2-1 - Rotating electrical machines Standard methods for determining losses and efficiency from tests Edition 1.0 2007.09 IEC 60034-2-2 - Rotating electrical machines - Standard methods for determining losses and efficiency from tests Supplement to IEC 60034-2-1 - Edition 1.0 2010.03 IEC/TS 60034-2-3 - Rotating electrical machines Specific test methods for determining losses and efficiency of converter-fed AC induction motors Edition 1.0 2013.11 IEC 60034-4 Rotating electrical machines Methods for determining synchronous machine quantities from test Edition 3 2008.05 IEEE Std. C50.12 IEEE Standard for Salient-Pole 50 Hz and 60 Hz Generators and Generator/Motors for Hydraulic Turbine Applications Rated 5 MVA and Above 2005 Page 23