Nordic Grid - FNR Frequency Containment

Transcription

1 Statnett SF Nordic Grid - FNR Frequency Containment Generating Equipment Performance - Review Report Assignment no.: Document no.: 01 Version: R-EXT

2 Client: Client s Contact Person: Consultant: Assignment Manager: Technical Advisor: Other Key Personnel: Statnett SF Norconsult AS, Nedre Fritzøegate 2, NO-3264 Larvik Terje Ellefsrød Terje Ellefsrød Einar Kobro, Hans Åke Glawing R2 R Reviosed after comments from workgroup Terje Ellefsrød Version Date Description Prepared by Checked by Approved by This document has been prepared by Norconsult AS as a part of the assignment identified in the document. Intellectual property rights to this document belongs to Norconsult AS. This document may only be used for the purpose stated in the contract between Norconsult AS and the client, and may not be copied or made available by other means or to a greater extent than the intended purpose requires. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 2 of 68

3 Summary The report identifies what type of hydro turbines and distributions between different types, number of units, operating heads and accumulated power associated to the Nordic Grid, limited to units > 10MW. Essential performance characteristics of the common turbine types are discussed including dead bands and backlash issues. Many of these parameters are not commonly discussed in the industry literature and must be considered to be presented as average expected values. Many are however supported by tests by Norconsult and others. The conclusion is that power dead band for the machine park overall can be expected to be about 0.5% weighed on MW installed but will have substantial variations from unit type to unit type. Because Pelton turbines in the system are employed with stability neutral governors (governors with a structure that retains an almost constant phase and gain margin in isolated operation regardless of droop setting) and have very moderate deadband and backlash, these turbines will most likely and according to test results - dominate in the corrective action related to 60sec swing around 30mHz amplitude. It appears quite likely that this mentioned pelton turbine control action is oscillating with units with backlash and time delays and short integral time that renders the response to be too late and too aggressively. Governor parameters tuning to lower the gain at the subject frequency to not exceed the dead band - would reduce the tendency to self- oscillate at this particular frequency but the phenomena would then probably shift to a lower frequency with higher amplitude. If HPC (parallel structure block diagram) with moderate transient gain and short integral time is employed to avoid penalty from dead band, Phase shifts will be substantial and most likely result in oscillating frequency. Influence from dead band and traditional governor tuning for HPC and Hyx governors is discussed for different droops and test amplitudes. Test amplitude <0.1Hz will be problematic. Discussion of BERTA test amplitude and bandwidth correlation to the dead bands is included. Finally, some suggestions to what may be done to improve or avoid further deterioration of the control performance has been made. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 3 of 68

4 Contents Introduction 8 Database 10 Turbines tested in FCR program 15 Francis turbines Guide vane rotation Friction in guide vane stem journals Sub-classification of Francis Turbines Francis turbines with conventional regulating ring Francis turbines without pressure regulating valves Francis turbines with pressure regulating valve Individual Guide Vane Control Time delays Improved control accuracy Medium size Francis turbines 125MW>Pn>10MW 27 Kaplan Turbines Kaplan Turbines guide vanes Runner Blade Control 31 Pelton Turbines 39 Servo Valve Pressure Gain 41 Solutions to Dead Band Power Feedback Discussion Dead band Compensation Mechanical Backlash and lost motion for specific groups Overall totalized dead bands Governor performance and tuning issues Practical implication of deadband Bandwidths participation Verification of small signal response for Pelton turbines 63 c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 4 of 68

5 Amplitude of test frequency 64 Likelihood of conformance with requirement characteristics 64 Improvements Address reduced basic performance of equipment over time Improvement control philosophy Improvements testing for certification References 67 c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 5 of 68

6 Error! No filename specified in document reference on page 1 Figure 1 Distribution of turbine capacity versus total MW and total unit count Figure 2 Mix of unit type in the Nordic System (% of installed MW in units >10MW) Figure 3 Distribution of Francis turbine capacity versus total Francis MW and total Francis unit count11 Figure 4 He (Design Head) distribution for Francis Turbines in the Nordic Grid Figure 5 Distribution of Pelton turbine capacity versus total Pelton MW and total Pelton unit count Figure 6 Distribution of Francis turbine MW per country Figure 7 Distribution of Kaplan turbine MW per country Figure 8 HPC 640 block diagram as applied in Dinorwig... Fel! Bokmärket är inte definierat. Figure 9 Control Loop applied in Dinorwig Verification Model... Fel! Bokmärket är inte definierat. Figure 10 Amplitude and Phase diagram from test of HPC 640 compared to modelling Fel! Bokmärket är inte definierat. Figure 11 Block diagram for the predominant types of Norwegian governors (T f =0.1T d ) Fel! Bokmärket är inte definierat. Figure 12 From Hymatek homepage... Fel! Bokmärket är inte definierat. Figure 13 Kaplan Turbine Load rejection HY10 governor. Delay <0.15sec, runner and guide vanes move simultaneously.... Fel! Bokmärket är inte definierat. Figure 14 Test result from 6GB4 governor controlling Kaplan turbine... Fel! Bokmärket är inte definierat. Figure 15 Approximate distribution of different governors controlling Norwegian generators Fel! Bokmärket är inte definierat. Figure 16 He 81 m 90MW New bushings c.o.f 0.05, worn Figure 17 He 400m Francis Turbine New Bushings c.o.f 0.11, worn state Figure 18 Guide Vane Force Indication H440m turbine, worn state (red) versus new state (blue) Figure 19 Indication of operating force, individually controlled guide vanes, turbine He 100m, 450MW after replacement of journal bearing - GV 15 & Figure 20 Indication of operating force, individually controlled guide vanes, turbine He 100m, 450MW after 18months operation following journal bearing replacement GV 15 & Figure 21 Cross Section He=500 m Francis turbine Figure 22 Cross Section He=150m Francis Turbine Figure 23 Cross section of guide vane system with 28 guide vanes He 495m 2008 design Figure 24 Calculated influence on discharge dead band (specific Francis units) from GV stem twisting c.o.f Figure 25 Graphic representation of tabular presentation of Guide Vane Twisting Figure 26 Example of guide vane mechanism with strong linkage rotation near closed position Figure 27 Example of typical high head guide vane control scheme with moderate linkage rotation Figure 28 Distribution of Kaplan Turbines versus number of Kaplan units and total Kaplan MW Figure 29 Distribution of Kaplan Turbines head versus number of Kaplan units and total Kaplan MW 29 Figure 30 KMW Kaplan 1985 (new measurement) - requires alternating GV servomotor force only at near full discharge. 6% of force capacity to overcome friction Figure 31 LMZ Kaplan GV servomotor force alternates across the full operating range. Approximately 10% of servomotor capacity is required to overcome friction Figure 32 New Kaplan 18m head GV operating force. Alternating force required for range <65% and >85%. Friction is about 10% of servomotor force capacity Figure 33 Measured GV servomotor by William Forsstrøm, Eksamensarbete 30hp force 10% of capacity to overcome friction, alternating force not required for normal operating range > 15 deg Figure 34 Operating force indication (Russian Kaplan 20m head turbine in Norwegian plant) friction 300kN of 1500kN = 20% - oil filled runner hub- backlash is not an issue for 70% and higher opening Figure 35 Nohab Kaplan 1965,33m head Friction 25% of servomotor capacity oil filled runner, approximately 1.5-2% lost motion is apparent from chart at positions above 6 deg Figure 36 From "Eksamensarbete 30hp" 2015 by William Forsström - Selsfors G1 He 22m Friction requires 25% of runner servomotor capacity oil free runner Figure 37 Kaplan with new runner (oil free) he 16m, friction requires 15% of the operating force. Lost motion about 0.5% average c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 6 of 68

7 Figure 38 Kaplan Turbine MELK ø6.0m 30 years operation conventional hub. Chart suggests 3% lost motion Figure 39 Commissioning test from 1985 (KWM) Figure 40 Load rejection 1985 KMW Kaplan Figure 41 Force indication Kaplan Turbine Runner blades (bronze bushing oil filled) Figure 42 Cross section of Pelton injector used from the early 60 s Figure 43 Servo Valve Pressure Gain Figure 44 Servo Valve Characteristic at Dinorwig Figure 45 Probable average power dead bands for hydro units in Nordic Grid Figure 46 Accumulated probable dead band for total number of installed Hydro units in Nordic Grid, distribution per unit installed Figure 47 Correlation between amplitude and steady state response for typical Norwegian Hyx and Vattenfall HPC tuning Droop 2% This chart is amplitude independent (no dead bands accounted for) Figure 48 Correlation between amplitude and steady state response for typical Norwegian Hyx and Vattenfall HPC tuning Droop 4% This chart is amplitude independent (no dead bands accounted for) Figure 49 Correlation between amplitude and steady state response for typical Norwegian Hyx and Vattenfall HPC tuning Droop 10% This chart is amplitude independent (no dead bands accounted for) Figure 50 Proposed Performance requirement Figure 51 Influence on response vs ω from governor (Hyx with standard settings, Droop 2%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.1Hz Figure 52 Influence on response vs ω from governor (Hyx with standard settings, Droop 2%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.05Hz Figure 53 Influence on response vs ω from governor (Hyx with standard settings, Droop 2%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.03Hz Figure 54 Influence on response vs ω from governor (Vattenfall HPC standard settings, Droop 2%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.1Hz Figure 55 Influence on response vs ω from governor (Vattenfall HPC standard settings, Droop 2%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.05Hz Figure 56 Influence on response vs ω from governor (Vattenfall HPC standard settings, Droop 2%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.03Hz Figure 57 Influence on response vs ω from governor (Hyx with standard settings, Droop 10%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.10Hz Figure 58 Influence on response vs ω from governor (Hyx with standard settings, Droop 10%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.05Hz Figure 59 Influence on response vs ω from governor (Hyx with standard settings, Droop 10%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.03Hz Figure 60 Influence on response vs ω from governor (Vattenfall HPC standard settings, Droop 10%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.1Hz Figure 61 Influence on response vs ω from governor (Vattenfall HPC standard settings, Droop 10%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.05Hz c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 7 of 68

8 Figure 62 Influence on response vs ω from governor (Vattenfall HPC standard settings, Droop 10%) connected to turbines with different position dead bands) compared to response for frequency amplitude 0.03Hz c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 8 of 68

9 1 Introduction Hydropower units provide a dominating portion of Frequency Containment Reserves [FCR] in the Nordic Grid. A series of frequency sweep tests of units carried out by producers and Gothia Power at various units indicates quite wide spread in response from unit to unit. To better understand FCR-N (Frequency Containment Reserve,Normal ) capability in the Nordic Grid, Norconsult has been retained to perform a desktop analysis of characteristic data of hydropower units in the system. The first task has been to update unit database Norconsult has been keeping for the Norwegian hydropower units to cater for changes (upgrades, decommissioned units, and new units) during the period from last update of the database (2000) In addition, a database of Swedish units has been created based on Kuhlin s plant database (Kuhlin, 2015) of Swedish hydro plants, broken down to unit level by detailed input from Norconsult experts in Sweden and the survey database from Nordic Grid (Data, 2015). Finnish units have been added based on survey data (Data, 2015) and available information posted by the major Finnish owners of hydro plants (Kemijoki, PVO, and Fortum). Overview of types of governing systems have been offered by Fortum Oyj, ans PVO Control Features of Hydro Turbines The review presented here was initiated as a response to temporary conclusions by the FCR Project team presented Sep , where dead bands (backlash) of to 0.006pu were introduced to non-linear simulation models to match observed oscillation. Power oscillations have assumed to be the root cause of the oscillations. The power oscillations is initially assumed to come from the load variations. This is now studied in a separate Imbalance study project. Tests performed indicate that not all units participating in frequency control damp these oscillations properly.. To better understand if specific unit characteristic inflicts higher or lower susceptibility to dead bands and dead times, this data collection and data analysis starts with a broad sorting of plants in Finland, Norway and Sweden into the typical unit type (Francis, Pelton, Kaplan), capacity [MW] and operating head (for Francis turbines. Review of the mechanical design of the significant types of turbines (Pelton, Francis and Kaplan) illustrates major differencies regarding probable dead bands between governor setpoint change and obtained MW change. This review suggests that high head Francis turbines such as what represents 40% of installed Francis capacity (MW) or 25% of the total hydro capacity will probably be associated with mechanical /hydraulic dead band 0.30% to 0.80% depending on many factors such as basic design and wear.). According to this study, Medium and low head Francis turbines will have less mechanical / hydraulic band, from 0.1% to 0.4% pending the wear and design features. Kaplan turbines, the second most dominating type of turbines in the system from a MW standpoint (23% of installed hydro capacity) will according the design features identified in this study be hampered with a mechanical backlash dead band related to the runner blade control. Runner blade positioning factors strongly into reaching desired power setpoint at medium and high discharges. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 9 of 68

10 Magnitudes of runner blade positioning dead bands have been evaluated based on common hub designs and based on this review often reach 1%. For older designs with control valve located in the hub the dead band will be higher. Guide vane stem twisting is in most cases an insignificant contributor to dead band on Kaplan units. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 10 of 68

11 2 Database The Norconsult database of hydropower units connected to the Nordic System contains a total of 45900MW of hydro turbines >10MW for Finland, Norway and Sweden. This amounts to a total of about 964 turbines. The database contains also small hydro units <10MW<1MW for Norway. Based on (Kuhlin, 2015) and Norconsult database, Approximately 5000MW is installed in hydro units <10MW. Such units are rarely equipped with a droop type governor and therefore disregarded in this review. Figure 1 Distribution of turbine capacity versus total MW and total unit count From this chart, it s evident that 35% of the installed capacity is concentrated in 10% of the units, or <100 units. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 11 of 68

12 Figure 2 Mix of unit type in the Nordic System (% of installed MW in units >10MW) Figure 3 Distribution of Francis turbine capacity versus total Francis MW and total Francis unit count c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 12 of 68

13 Figure 4 He (Design Head) distribution for Francis Turbines in the Nordic Grid Figure 5 Distribution of Kaplan turbine capacity versus total Kaplan MW and total Kaplan unit count c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 13 of 68

14 Figure 6 Distribution of Pelton turbine capacity versus total Pelton MW and total Pelton unit count Figure 7 Distribution of Francis turbine MW per country c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 14 of 68

15 Figure 8 Distribution of Kaplan turbine MW per country. Figure 2 and Figure 4 show that high head Francis turbines makes up 0.6 x 0.4= 24% of the hydro capacity in the Nordic countries and Kaplan turbines 23%, total 47%. This is a significant observation since the dead bands for these types of units as discussed in Chapter 4 and 5 and summarized derived throughout the discussions and presented in 9.2 will render such units providing insignificant control input for the amplitude 0.03Hz at 60s period frequency illustrated in Figure 52 to Figure 63 Pelton turbines are exclusively installed in Norway Since the distribution of units is spread out among a wide range of unit capacity ranges, many different technologies are applied. When also taking into account that the three dominating unit types (Francis, Kaplan and Pelton) are in use, it s likely to find unique issues related to frequency control, partially because of how the equipment is constructed and partially because of features of the water conduits they are associated with. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 15 of 68

16 3 Turbines tested in FCR program A series of tests with Hardware in the loop were carried out by Gothia power (Power, 2014). These tests involves disconnecting the frequency measurement input to the governor and replacing it with a synthetic frequency source with variable amplitude and where oscillations at any frequency and amplitude can be overlaid the common 50Hz signal. Response from the unit is derived by measuring the power fed to the grid. The measured power will therefore include inertial effect from the generator s speed change as response to actual power grid frequency fluctuations at the time of measurement. It can be expected that this error grows with lower frequency oscillation amplitude. At frequency ampliutudes mostly applied, 0.10Hz and above, such effect is not significant. Name Turbine Type Governor Type Guide vane control Head [m] Power [MW] S1 Kaplan Regulating ring S2 G1 Kaplan Standard Vattenfall **** S3 Kaplan Deriaz Standard Vattenfall **** S4 G1 Francis Standard Vattenfall **** S5 Francis Standard Vattenfall **** F1 F2 F3 Kaplan Vertical Kaplan Vertical Kaplan Vertical Regulating ring Individual*** Individual*** Regulating ring Kejo Regulating ring KTR-2102 Regulating ring Regulating ring N1 Pelton 5 Jet Hymatek1x Individual* N2 Pelton 4 Jet Voith 6GB94 Individual* N3 Pelton 6 Jet Kværner TC210 Individual* N4 Pelton 5 jet Hymatek1x Individual* N5 Francis Hymatek1x Regulating ring N6 Francis Andritz 1703 Regulating ring *individual injector control, mechanical pilot **individual injector control, electronic***fieldbus control, **** Standard Vattenfall is similar to standard Swedish. Kaplan, Francis and multi jet pelton turbine types have been tested. S3 is a typical since this is the only Deriaz turbine in the system. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 16 of 68

17 4 Francis turbines Francis turbines power 60% of the installed Nordic hydropower capacity. Common head range for Francis turbines is 50 to 600m, the distribution according to unit count and MW accumulated is presented on Figure 4. The 50 percentile point (average weighed on Unit capacity) is 230m, based on unit count the average head is 150m. The 10% (about 45turbines) highest capacity Francis turbines Pe>125MW represents 35% of installed Francis MW capacity, making it potentially a stereotype community of units to be reviewed. 4.1 Guide vane rotation Flow control and therefore power control for Francis turbines is by rotating the guide vanes by twisting the guide vane stems. For low specific speed high head Francis turbines, the guide vane needs to be rotated around degrees from speed no load to full rated load. For Francis turbine with high specific speed (low operating head, around 60m for medium and large size units) the guide vanes must be twisted about 25 degrees from speed no load to rated power. Values can be interpolated. The operating lever that in turn is connected to the linkage and the operating lever are attached to the top of the stem. This stem will twist to transfer the torque Friction in guide vane stem journals Without getting into the factors that determines the stem diameter, it is clear that from a regulating accuracy point of view, the shaft will have to twist to overcome friction that occurs at the opposite end of where the regulating arm is attached to the link (and regulating ring). Friction will always oppose the motion and hence the twisting is not affected by changes absolute torque direction, only the difference in torque. Examples of friction in the guide vane systems when new and after years of operation are not commonly available in the literature. Therefore, Norconsult in-house test data for operating pressure in servomotors during operation or test data made available to Norconsult have been used based on. Four examples: Figure 9 Francis Unit He 81m, 90 MW (increase in friction 4x 6x after 30 years Figure 10 Francis Unit He 400m, 125MW (increase in friction 2x after 30 year Figure 11 Francis Unit He 500m 320 MW (increase in friction x after 25 years Figure 12 Francis Unit He 100m 450MW (increase 1.8 x after 1.5 years) Each of the figures charts the force that is derived from measured oil pressure at closing side and opening side of the guide vane servomotors as a function of the position of the guide vane. The log can be taken manually while moving the guide vanes slowly or automatically with a logging system recording at frequency about 10Hz or higher opening pressure, closing pressure and servomotor position. All measurements are made while unit is in service connected to the grid. If there were no c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 17 of 68

18 friction, force when guide vanes are moving towards open would be identical to force measured when guide vanes are closing. Different force at the same position when guides vanes are moved in alternate directions is due to friction force. The friction is acting opposite to the direction of motion. Therefore, the difference between the force when moving towards open and the force when moving closed is 2 x the friction force. Figure 9 He 81 m 90MW New bushings c.o.f 0.05, worn Figure 10 He 400m Francis Turbine New Bushings c.o.f 0.11, worn state 0.22 c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 18 of 68

19 Figure 11 Guide Vane Force Indication H440m turbine, worn state (red) versus new state (blue). Figure 12 Indication of operating force, individually controlled guide vanes, turbine He 100m, 450MW after replacement of journal bearing - GV 15 & 20 c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 19 of 68

20 Figure 13 Indication of operating force, individually controlled guide vanes, turbine He 100m, 450MW after 18months operation following journal bearing replacement GV 15 & 20 To calculate the absolute coefficient of friction, the guide vane lever, arm, regulating ring and servomotor configuration must be taken into account. Further, the friction associated to the regulating ring and servomotor should be deducted. This friction is a function of th regulating ring s weight and bearing arrangement. The weight of the regulating ring is generally a proportional to the servomotor force capacity. The coefficient of friction (c.o.f) can be assumed similar to the guide van stems. Hence, the friction that is associated to the regulating ring corresponds to a certain percentage of the servomotor force capacity. Based on this approximation, it has been established that regulating ring friction corresponds to 0.75% of servomotor force capacity. Single servomotors controlling the guide regulating ring and twin servomotors with one servomotor expanding and one withtracting when the regulating ring moves, result some times in imbalanced force pairs. This force imbalance must also be taken into account since it s influence will skew the apparent guide vane stem friction. The radial force in the journals of the guide vanes caused by water factors into the calculation of friction torque, (NTNU, 2006), Figure 14. The factors Ω and α denotes empirical factors pending the specific speed and angle of the guide vane position respectively to make up the specific radial force. The empirical data can be derived from evaluation of pressure in the vanless space from model tests, from prototype tests or from CFD calculations. For calculations in this report pressure in vaneless space from model tests for one turbine of medium head (150m) extrapolated to higher and material from lower heads around guide vane position fro b.e.p has been used and extrapolated. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 20 of 68

21 Figure 14 Calculation of friction torque for guide vanes (NTNU, 2006) For the calculations, the guide vane ring pitch diameter, D 0 the height of the guide vanes (B 0 ), the stem diameter (d) and number of guide vanes have been accounted for according to the specific turbines design drawings. Also of importance when calculating the dead band is the length of the guide vane stem and the angle guides vanes are turned from speed no load to full load. This is also an empirical correlation, presented on page 17 (NTNU, 2006). By taking such measurements from observations made on 4 Francis turbines when new and after operating for years it has been possible to obtain a better idea of the guide vane stem friction coefficients and to a certain extent the development of the friction with time. Friction coefficients are expected to range from 0.06 for new greaseless bushings to 0.25 for worn systems. This is well within published values from manufacturers and dedicated friction measurements to mechanical systems - old and new - that Norconsult do. It is possible to break down servomotor force measurements to derive coefficient of friction, albeit with some level of uncertainty margin. The dominant uncertainty According to the measurement data and calculation according The lifetime span for friction to increase appears to be q bit shorter than what is commonly used for overhaul interval but the statistically available data is insufficient for reliable conclusions. An average coefficient 0.15 for the overall population of Francis turbines with time from overhaul spanning from 1 to 40years seems like a sensible but slightly optimistic value. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 21 of 68

22 Figure 15 Cross Section He=500 m Francis turbine Figure 16 Cross Section He=150m Francis Turbine 2010 c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 22 of 68

23 Figure 17 Cross section of guide vane system with 28 guide vanes He 495m 2008 design Among the group of turbines (Large Capacity Francis Turbines) making up 35% of the overall dominant Francis turbine population, 55% of the capacity originates from the Norwegian high head Francis units with H>300m. With average coefficient of friction assumed to be 0.15, the stem twisting invoked dead band will most likely be affected by the design head as shown in Figure 18 and Figure 19. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 23 of 68

24 Pnom [MW] Head [m] Installed Yr Dead Band ±[%] Figure 18 Calculated influence on discharge dead band (specific Francis units) from GV stem twisting c.o.f 0.15 Figure 19 Graphic representation of tabular presentation of Guide Vane Twisting This review implies that dead band from twisting of the stems is quite pronounced with high head Francis turbines with all units reviewed with 400m and higher head yielding between 0.2 and 0.3% dead band from this source while units with 120m and lower head yields <0.10% dead band from this source. Figure 4 show that 27% of the Francis turbine capacity in the Nordic grid originates from units with head over 400m, 12% from units with head above 500m. A theoretical relative deflection of about double magnitude compared to comparable units from 1980 s and earlier appears for designs 2008 and later due to lower stiffness in the guide vane stem relative to the friction torque. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 24 of 68

25 4.1.2 Sub-classification of Francis Turbines The population of Francis turbines above 125MW represents 35% of the Francis turbine capacity and represents potentially stereotype group, with two distinct classes of units: Medium head m large units in Sweden (about 3900MW) High head high capacity in Norway (units >approximately 300m head) (about 5000MW) Two of medium head turbines have been tested in Sweden. S5 G1 and S4 G1 N6 (also a tested unit) in Norway is near the classification with head and output that is marginally lower but with technical design features that matches many of the units inside the classification. 4.2 Francis turbines with conventional regulating ring A common class of Francis turbines features guide vanes (20-28psc) controlled by 1 or more servomotors rotating a regulating ring that in turn is connected to guide vane stems by individual linkages to arms on the guide vane stems. Beside discharge capacity and head, and type of guide vane control system, there are two other classifications of Francis turbines; with and without pressure regulating valves Francis turbines without pressure regulating valves The majority of medium and low head Francis turbines have no pressure regulating valve. Within the class discussed here, Francis turbines > approximately 125MW, 85% or about 8000MW has no pressure regulating valve (Norconsult, 2015). For turbines without pressure regulating valve, the guide vane linkage geometry is typically arranged with a strong opening tendency near closed position. There are various reasons for this. Abrupt closing off of the flow near zero flow condition may result in waterhammer. By arranging linkages and arms with strong rotation tendency of the link near closed position, the rate of changing flow at a given motion of the servomotor can be reduced by as much as factor 3 compared to flow change around 50-70% opening,, see Figure 20. The linkage geometry above also increases the available torque that can be transferred near closed position for a given servomotor capacity, albeit at slightly longer total servomotor stroke than would otherwise be necessary In a scenario with loss of oil, theoretically the guide vanes will not slam totally shut but remain slightly open, avoiding high overspeed if disconnected from the grid while avoiding high waterhammer as would potentially be the case if guide vanes were self-closing also near closed position. Generally, these types of turbines will for moderate wear rates and moderate friction not require alternating force from the guide vane servomotors in the normal operating range. This will reduce the impact from backlash. However, if c.o.f increases to more than impact from backlash may become an issue for this type of turbine according to common layout principles for the guide vanes. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 25 of 68

26 Figure 20 Example of guide vane mechanism with strong linkage rotation near closed position Francis turbines with pressure regulating valve Among the larger capacity high head Francis turbines in Norway in the category discussed here, about 1000MW (12% of the segment s population) includes or was originally designed with a pressure relief valve to allow guide vanes to close fast. This reduces transient overspeed and pressure for large load throw off. The pressure relief valve will not affect optimal governor tuning for small signal stability. There is, however, a tendency for the guide vanes for turbines with pressure relief valves to be balanced closely around neutral water torque for wide operating ranges, see Figure 10 and.figure 11 The main rational for different balancing of the guide vanes in these scenarios with pressure relief valve is improved linearity between guide vane angle ( and water discharge) and servomotor positon since servomotor displaced oil is directly attracting pressure relief valve motion. Strong un-linearity would tend to reduce the effectiveness of a pressure regulating valve. Minimizing un-linearity is thus prioritized. This requires smaller operating forces overall and makes alternating force requirement from the servomotor more likely even with moderate friction. Linkage backlash will be triggered if alternating force is required. Medium and large size high head Francis turbines where pressure regulating valves are most common have lever motion in the range of 300mm in the active load range. With journal diametrical clearance in each end of the linkage as new 0.25mm, the lost motion can amount to (0.5/300)=±0.08% as new and ±0.16% in a worn state. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 26 of 68

27 Figure 21 Example of typical high head guide vane control scheme with moderate linkage rotation 4.3 Individual Guide Vane Control A relevant design feature of the high capacity Francis turbines as it relates to frequency control is related to applications with individual guide vane control, without regulating ring. About 2000MW Francis turbines in the system features individual guide vane controls without a regulating ring, all in Sweden and all >120MW. S4 (one among the tested turbines) features this type of control. At S6 G5 (450MW) the OEM hydraulic system with mechanically linked multiple control valve system is still in place. S6 G5 is also the newest of the individual control systems. At all other sites, the mechanically linked system has been replaced with individual proportional valves for each guide vane servomotor Time delays Based on test results in S4 featuring modernized control systems for the guide vane servomotors, the guide vane controls are associated with delays of about msec from sources in the digital position controller system that is unique to this type of system. Delays introduced by cycle time in the Fieldbus system have been reported by Waplans ( (Waplans, 2001) related to S3 conversion) to about 130ms. In addition there is a servo positioner cycle time of 30 ms. Time delays in the speed governor itself is not accounted for here. Gothia Power modelling of S3 with baseline in recorded guide vane position equalized time constant of 600msec (not modelled as time delay) appeared to match measurements well. However, this test was applied with very high frequency oscillation amplitude (0.5Hz) and 20% droop and with more realistic values time constant and time delays would most likely worsen. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 27 of 68

28 It seems therefore likely based on the tests and published material that individually controlled guide vane systems with modernized fieldbus control will be associated with abnormal time delays. Only S6 G5 remains in service with the original analogue mechanical hydraulic servo control.. All other individual guide vane control systems will have high likelihood of serial bus related time delays Improved control accuracy Important arguments for the individual guide vane control when introduced in 1965 included better frequency control responses (Nohab, 1970). This improvement was in particular argued for the wing servomotor scheme. Also contributing to smaller dead bands would be the uni-directional guide vane torque where even worn joints would not result in backlash. Finally, the regulating ring by itself has journals and supports that become worn. Elimination of the regulating ring and linkages was also considered an improvement, in particular in the region of operation where water hydraulic forces are close to neutral. All wing servomotors have been replaced by conventional individual servomotors today. 4.4 Medium size Francis turbines 125MW>Pn>10MW Medium size Francis turbines (<125MW approx.) in the Nordic System are always controlled by one or more servomotors and a regulating ring. There are units of this size range across the entire range of heads that are common for Francis turbines. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 28 of 68

29 5 Kaplan Turbines The second most dominant group of units in the Nordic system based on accumulated MW are the Kaplan turbines. Vertical Kaplan turbines are generally applied for heads 10 45m and horizontal Kaplan bulb turbines from 4 15m head. The flow control and thus power control on Kaplan turbines combine moving guide vanes and runner blades. Individual control of guide vanes is applied for some large Kaplan bulb turbines. The contribution of bulb turbines in the FCR-N picture is however insignificant. 35% of the overall Kaplan turbine capacity is installed in units > 50MW. The average head of Kaplan turbines weighed on MW is 23m. Figure 22 Distribution of Kaplan Turbines versus number of Kaplan units and total Kaplan MW c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 29 of 68

30 Figure 23 Distribution of Kaplan Turbines head versus number of Kaplan units and total Kaplan MW Two Kaplan turbines in Sweden and one in Norway feature individual guide vane servomotors. This applies to the largest Kaplan unit in the system, 180MW S7 in Sweden, 150 MW S8 and Norwegian N7 110MW. S7 and N7 uses mechanically control guide vane valves while S8 uses digital bus controlled individual valves. S8 U1 is in the process of being replaced by a new plant and it can thus be said that Kaplan turbines with individually controlled guide vanes is not a significant population, representing less than3% of the installed Kaplan capacity. 5.1 Kaplan Turbines guide vanes The guide vanes are in a similar fashion as Francis turbines twisted to guide water onto the turbine. The influence from twisting the stem is, however, much smaller than for Francis turbines because of lower head, tendency of larger margins between normal operating torque and design torque for off cam conditions because of Kaplan characteristics and shorter stem relatively seen. Dead band caused by guide vanes stem twisting will be 0.01% to 0.03% in most cases. Figure 24-Figure 27 illustrate guide vane force indication for 4 Kaplan turbines ranging in head from 13-20m and output MW. Of the 4 units, servomotors for 2 units can position the guide vanes without alternating the force in the common 70%-80% opening range. The two other turbines will be exposed to backlash in the linkage system. (positions is not exposed to backlash in the guide vane control system c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 30 of 68

31 Figure 24 KMW Kaplan installed and tested 1985 ( As new measurement) - requires alternating GV servomotor force only at near full discharge. 6% of force capacity to overcome friction. Figure 25 LMZ Kaplan GV servomotor force alternates across the full operating range. Approximately 10% of servomotor capacity is required to overcome friction. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 31 of 68

32 Figure 26 New Kaplan 18m head GV operating force. Alternating force required for range <65% and >85%. Friction is about 10% of servomotor force capacity. Figure 27 Measured GV servomotor force (Forsström, 2015) 10% of capacity to overcome friction, alternating force not required for normal operating range > 15 deg 5.2 Runner Blade Control There are two main variations of the runner blade controls: a) The main hydraulic regulating valve controlling flow ported installed external to the turbine, porting oil through borings or pipes through the shaft centreline to the opening side and closing side c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 32 of 68

33 b) A pilot servo valve controlled first stage in turn controlling the position of a valve internal to the rotating shaft. This valve is distributing governor oil to either side of the piston. There is no actual feedback of the blade position for this system, only the set point (position of the control valve) (Hansson, 1977) writes In order to improve the accessibility of the control valve of the runner servomotor this has been moved to the combinator on top of the unit or in certain cases been placed outside the rotating system. This has also brought about a higher precision in the governing and a more reliable indication of the actual position of the runner blades. Gradually, upgrade projects retires the hub mounted control valves of the 60 s and earlier Kaplan turbines. The position dead band incurred from the internal valve is hard to estimate. But around 1.0% is a fair estimate. Very high accuracy of runner blade positioning has not commonly been a design objective, moderate oil consumption in the runner blade control system has been the highest priority. Figure 34 shows positon dead band is about 1º or 3%. runner blade position for a 1985 issue large Kaplan turbine According to information received from a Finnish producer, their machine park includes 27 Kaplan units up to 50 MW 10 renovated high pressure units 11 low pressure units where runner blades main relay valve is located inside the rotating shaft 6 low pressure units, where main control valve is outside of the rotating shaft. 11units means 40% when based on unit count. From the producer data summary 4 of 17 Kaplan turbines employ internal control valve, corresponding to 26% of the installed capacity. The subject units were Tampella units installed in Survey data from another Finnish producer(19 Kaplan turbines, 970MW) shows that 10 units, 370MW or 40% of the turbine capacity uses control valve in the hub and no direct feedback of blade position. It would be a fair assumption that most Tampella units installed prior to 1970 have not been converted to oil free runner hubs feature this type of runner blade control. Swedish data in this respect has not been available but it is a fair assumption that the situation is similar to Finland, with 30-40% of the installed Kaplan power with runner blade control without direct feedback. This type of mechanisms are gradually being replaced when Kaplan hubs are replaced. But it is a relatively slow development. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 33 of 68

34 Figure 28 Kaplan hub internal parts, typical layout applicable for most units Figure 29 to Figure 33 illustrates indication of operating force for 5 Kaplan turbine runner (hubs). 3 of the hubs require alternating force from the servomotor. Figure 34 shows the backlash reported in the commissioning report (KWM 1985) Figure 35 shows load rejection for the same KMW Kaplan turbine (1985) where runner closing time constant / delay is readily visible. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 34 of 68

35 Figure 29 Operating force indication (Russian Kaplan 20m head turbine in Norwegian plant) friction 300kN of 1500kN = 20% - oil filled runner hub- backlash is not an issue for 70% and higher opening. Figure 30 Nohab Kaplan 1965,33m head Friction 25% of servomotor capacity oil filled runner, approximately 1.5-2% lost motion is apparent from chart at positions above 6 deg.- c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 35 of 68

36 Figure 31 Selsfors G1 He 22m Friction requires 25% of runner servomotor capacity oil free runner (Forsström, 2015) Figure 32 Kaplan with new runner (oil free) he 16m, friction requires 15% of the operating force. Lost motion about 0.5% average (Norconsult Archive) c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 36 of 68

37 Figure 33 Kaplan Turbine MELK ø6.0m 30 years operation conventional hub. Black line represents 0 force. Step in position with limited force pick up around 0 force suggests 3% backlash lost motion (Erich Wurm, 2013) Figure 34 Commissioning test from 1985 (KMW) c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 37 of 68

38 Figure 35 Load rejection 1985 KMW Kaplan Figure 36 Force indication Kaplan Turbine Runner blades (bronze bushing oil filled) The friction requires generally 15% to 30% of the capacity of the hydraulic piston. An often used principle is that the friction well exceeds the waterhydraulic force and demands a pulling force to close and pushing force to close. Or vice versa. In either case, all gaps in the two journals on c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 38 of 68

39 the linkage connecting the crosshead to the blade lever will have to be made up to effectively twist the blade. One out of six (16%) of tested units in our archive do, however, show that the design of the blades will cause no dominant influence from hub linkage wear since torque direction to position blades do not alter when blades are controlled in the generally most used range 60% and higher. The travel of the piston is typically around 150mm for the average Kaplan hub. Expected typical as new linkage clearance is around 0.25mm, in worn state around 0.5mm. The total lost motion from the linkage alone will thus be from ±0.25 to ±0.5mm or 0.5/150. The links are further tilted about degrees relative to the motion of the piston, affecting the relative lost motion by a factor 1.4 on the lower link. Total expected lost motion from the linkage system is therefore from new (0.125* ) = ±0.35mm to common worn level ±0.25* =±0.7mm. For a typical system this will result in lost motion dead band 0.35%-0.7% in the runner blade positioning for most Kaplan hubs. Analysis of test results included in this report suggests that the lost motion in the hub is in reality larger, maybe because hub wear in general is allowed to grow well above common assumption of double clearance or/and that wear of crosshead guides and blade trunnions journals influences the backlash also. The effective influence on power dead band from a runner backlash will incur a reduced MW response, not a total absence of response within the runner dead band since partial response will be obtained from guide vane motion. But the total discharge and power response requires both guide vane AND power to respond. Typical correlations around best efficiency point: a is guide vane pos, α is rummer blade position and Q is flow, all values pu. dq/dα= dq/da= It means that the response to a linearized flow discharge request change will respond to about 50% according to the guide vane backlash and the rest according to runner backlash. At the same time, however, the efficiency will typical drop 0.4%- 0.6% per 1% off cam guide vane position and the power response will only be 25%. The runner blade motion is responsible for the remaining 75% power. Statistical significance Results from 6 tested units are used to qualify general characteristics of the Kaplan population regarding the units under one (not affected by backlash) or another (affected by backlash) category. This is a so called Bernoulli distribution (Pishro-Nik, 2014). According to this theory, p is1/6 and hence Var X=0.134 in other words there is a reasonable likelihood that between ( = 0.026= 2.6%) and = (29.4%) of the Kaplan turbines will not be affected by backlash. For the purpose of this presentation where we are presenting characteristic tendencies and the general lack of availability of such test data, the uncertainty is large but the tendency is still clear. Backlash in Kaplan hubs is a potentially significant issue. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 39 of 68

40 6 Pelton Turbines The Pelton turbines make up about 17% of the Nordic grid hydropower capacity. Pelton turbines in the Nordic grid are exclusively found in Norway. The Pelton turbine uses oil hydraulically controlled injector needle valves to control the water free jet discharge onto the runner. There are two common control systems, each used on about 50% (Norconsult, 2015) of the installed Pelton turbine capacity: 1. Primary injector control where each injector has its own electro-hydraulic servo valve, with deflectors controlled over separate control valve(s) 2. Primary deflector control where the deflector position controls hydromechanical servovalves that for each needle. For this system there is three series positioning loops; a pilot stage, a main stage and 2-6 parallel injector stages. There are no mechanical links associated to transfer of the main control force on modern Pelton turbines, the oil hydraulic piston is only dm s from the jet needle. Figure 37 Cross section of Pelton injector used from the early 60 s Friction in the system will typically be within 10% of the available oil hydraulic force. The servo gain is often high (typically >30), resulting in very modest dead band associated to servo valve pressure gain. A specialty that may hamper control dynamics at varying degree is the low oil temperature of the oil that is residing in the servomotor, often surrounded by very cold water. The high viscosity oil resulting will reduce performance of valves and incur increased losses in oil pipes. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 40 of 68

41 Injector needle control valves on category 2 type systems has often lapping that results in injector needle control time constant of about 1-1.5s for position errors <1% Injectors for 4 6 jet Pelton turbines can typically be taken out of use (closed) for part load operation. Pending logics applied in the governor, this may or may not inflict noticeable change to the transient response. Needle selection control algorithms will most likely influence part load small signal PID response for 50%-70% of the Pelton capacity. Strong un-linearity ratio of about 2-4 from no load to full load between discharge and injector needle position on most Pelton turbines is predominant in particular for units with primary needle control. Linearization is normally incorporated for feed forward control but PID control is predominantly adapted without linearization. HY1x governors can set PID linearization by selecting ON/OFF parameter locally at the governor panel. HY1x governors control 25% of the Pelton turbine even though this function is normally deactivated capacity. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 41 of 68

42 7 Servo Valve Pressure Gain The servo valve is generally controlled by the governor via a closed loop positioning system. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 42 of 68

43 10% ΔP=0.3%U Figure 38 Servo Valve Flow Characteristic and Pressure Gain (Rexroth, 2009) c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 43 of 68

44 The servo valves features a pressure gain, it typically is similar to Figure 38 for the high response type valves. What this means is that friction in the controlled system will invoke a dead band caused by the servo loop pressure gain characteristic. With friction force 10% of the system capacity and servo loop gain is 10 (this is a common value for standard sized Francis and Kaplan guide vane control valves) according to the authors experience and also referenced in (R&D, 2015) the dead band invoked will be about 0.03%. Smaller requests for position shift will not activate motion. For larger friction, the dead band will grow proportionally. For Kaplan hubs, one can expect friction to account for 0.1% lost motion or more because of valve pressure gain characteristic. It is common to use less gain than 10 on the Kaplan runner control loop. It is also possible that some units employ servo valves with positive lapping to lower oil consumption. It invokes dead band, such as reported for Dinorwig where positive lapping is 2/60=3% (see Figure 39). With servo loop gain 10 it will cause dead band 0.3% in addition to dead band related to pressure gain. Examples are for instance if the servo valve s capacity is such that it administrates closing from open within 4 sec - or rate of motion 25%/sec, - the valve s controller can normally not use gain much higher than 10, in order for the servo loop to remain stable. If the servo valve s overlap is 1%, the postion dead band logically will become 0.1% since position errors are multiplied with the gain. There is generally a wide range of electro-hydraulic proportional valves in use. Manufactured by 3-4 main international supplier such as Bosch Rexroth, Vickers, Moog, Atos c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 44 of 68

45 Figure 39 Servo Valve Characteristic at Dinorwig (German Ardul Munoz-Hernandez, 2012) The band widths, dead bands and linearity, three important factors for the main control valve functional features, cannot be predicted with any accuracy out detail review of the exact model number of the valve as well as the application it is used in. Valve data presented from PVO (Fel! Hittar inte referenskälla.) and Kemijoki Fel! Hittar inte referenskälla. describe predominantly 4WRLE.. M characteristic, which is a fine metered characteristic according to Figure 38. BOSCH Valve is also used in the Finnish system (Kemijoki) and is equivalent to 4WRL. Bosch /061 are high bandwidth valves used by Kemijoki for pilot control of 2 stage systems. Test results from our files indicate that servo loop time constant less than 0.30 sec can be expected for high pressure systems with the 4WRLE or equivalent performance valve connected to 10ms cycle time digital controller (Figure 42) With an assumption that Hydraulic control valves in standard Swedish system also conform approximately to what is used in Finland and Norway, reported characteristics (R&D, 2015), time delays in the order of magnitude seen for S2, S3, S5 and S4 imply that the digital processing is hampered with time delays that are long enough to cause measureable negative influence on the performance. Figure 40 c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 45 of 68

46 S2 S3 S5 S4 Figure 40 Table of derived time delays, dead bands, and time constants c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 46 of 68

47 Figure 41 Valve spool positioning bandwidth characteristics for commonly used 4WRL (Rexroth, 2009) c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 47 of 68

48 Figure 42 Step test with 150bar hydraulic system, 4WRLE16V 200M valve, Servo Loop gain 35, Ty approx Figure 43 Step test with 150bar hydraulic system 4WRLE25 V370M, Servo loop gain 10, Ty approx. 0.25s Older systems that have two stage control valves have considerable positive valve lapping resulting in valve dead band similar to. This is often compensated for by introducing a dither in the hydraulic control circuit of the control valve at a frequency higher than the actual bandwidth of the main hydraulic system. c:\users\anwe\downloads\rep_nordic grid - fnr freq-rext.docx Page 48 of 68