| The following papers have been accepted for the IGHEM '98 Conference in Reno: |
This paper presents an indepth analysis of uncertainties in the measurement of the parameters used in the calculation of turbine efficiency by the Thermodynamic Method. Techniques for the installation and calibration of sensors are presented to minimize systematic errors. The location of temperature probes is considered in detail, including test results from a field investigation of the temperature and velocity profiles in the discharge from a large Francis turbine.
Performance tests were conducted at Unit 22 at the Société d'Ènergie de la Baie James' Laforge-2 plant between June 11 and 15, 1997. These tests included measurements of the discharge through the turbine using current meters. Simultaneous measurements were also taken in one bay of the intake with an Acoustic Scintillation Flowmeter (ASFM). The ASFM is a new instrument, which offers unique advantages for measuring intake flow in low-head, short intake plants for which current meters have been the traditional and only effective method. It is non-intrusive, and its deployment in intake gate slots is straightforward, allowing data to be collected with a minimum of plant down-time. LaForge-2 is typical of large to medium-sized plants of that type; it is equipped with two 147 MW Kaplan turbines, each with a three-bay intake. The bays at the metering section are 19.2m high and 6.1m wide. The net head for the plant is 27.4m.
The current metering used on hundred ninety measuring points in each bay, obtained using forty individual current meters mounted in four rows of 10 on a frame 4.6m high. The current meter rows were spaced 1.08m apart vertically. The inclination of the meters was controlled by a hydraulic adjustment system to align them with the flow. The ASFM transducer arrays were mounted on the same frame as the current meters in Bay 1, at the trailing (downstream) edge of the frame. Flow measurements were taken simultaneously with the current meters and the ASFM at a series of unit operating conditions. Measurements were taken using two different profiling methods: one with the frame at five fixed positions, using data from all four rows, the other with the frame profiling continuously, using data from the lowest row on the frame.
The trash racks had been removed from the intakes for the testing, resulting in low levels of turbulence in the flow. The low turbulence, combined with interference from the current meters mounted ahead of the ASFM transducers, hampered the performance of the ASFM when the fixed-position measurements were taken. The interference was least at the lowest array pair, which allowed meaningful data to be obtained during the profiling runs. The discharge through the bay was independently computed for thirteen cases (between 145 and 200 m3/sec), and the results then compared. Over this range the mean difference between the measurements was less than 0.7%. An analysis of the differences and the uncertainties in both methods is made. The results of the analysis are used to assess the ASFM's advantages for measuring turbine flows in low-head plants.
To determine the efficiency of a unit in a low head power plant, the current - meter method is almost the only suitable to meet a good uncertainty. The difficulties of measurement come normally from the oblique flow, the small length of the penstocks and the proximity of the trash racks. This article show the results of the discharge measurement at the La Grande-1 and Laforge-2 power plants.
Although the number of points for velocity measurement met the recommendation of the IEC code, the effect of the trash racks have been to produce a systematic error in the discharge at La Grande-1. The measurement was done by placing the current-meter frame at a first elevation to obtain the velocities for a portion of the section. Five more elevations were necessary to cover the entire section.
To prevent the effect of the trash racks at Laforge-2, the discharge measurement has been performed by sweeping the frame supporting the current - meters along the metering section. A special mechanism has been used to align the current - meters with the flow while sweeping.
This method gave results that agree very well with and without the trash racks. In addition, the method saved time of execution.
During an R&D program, comparative tests have been performed at the Laforge - 1 power plant in the James Bay project. Pressure - time, current-meter, acoustic and thermodynamic methods were performed with different modifications to our normal procedure and equipments. One of these was to lower partially the head gate to improve the mixing of the water for the thermodynamic method. The later was done to measure the efficiency of a unit at a very much lower head than recommended by the IEC code.
At the same time, the discharge measurement was performed with the current - meter method. With a frame not so far from an elbow, it has shown that the velocity profile has been greatly modified by the action of the head gate. The general effect was to smooth the velocity profile that look more like a measurement that would have been at many diameters downstream from the actual position. The smooth velocity profile can then lead a better uncertainty. The procedure can also have benefit effects for the other methods.
When the structural strength of the head gate is sufficient to support the static and dynamic lead, the measurement with and without the gate lowered can show an improvement in the flow condition, specially near a disturbance.
The paper details both thermodynamic and conventional tests carried out on 2-4MW pumps.
The tests were carried out independently on the Sulzer Test Rig at their Leeds Plant.
The traditional manner to measure the inlet temperature of a turbine is to use sampling probe with withdraws a small sample of the water at ( 1/7( from the liner. This sampling probe possesses an upstream oriented tap to capture the kinetic velocity as well as the static pressure. The IEC 41 publication does not impose the use of sampling probe, thus permits to place the thermometer directly in the main flow. Then arises the question: how the kinetic energy is converted into temperature and what should be the design of the thermometer sensor to do it correctly.
At the first IGHEM seminar in Montreal, Mr. Pederson of SBE presented and paper showing the importance of the velocity effect on the temperature. Few years ago, the author of the present has proposed to measure the temperature directly in the main flow and has presented a paper titled 'Mode opÈratoire par Immersion pour la mÈthode thermodynamique', in Chamoinix, France. Since industrial and rugged thermometer of high accuracy now exists, the effect of kinetic velocity on temperature is of great concern especially for large machines.
A test was done in a power plant having a very long power tunnel and 4 thermometers at the turbine inlet were used. Two were placed directly in the main flow and the two others with sampling probes. The thermometers used were calibrated at SBE shortly before the test and the site measurements demonstrated that the water homogeneity was of 1 mK so as the accuracy of the measurements. The measurements were done for water flow velocity ranging from 1 to 9 meters per second. On a 200 m head turbine, 2% on the efficiency represent the equivalent kinetic energy than a velocity of 9 meters per second. So, this effect of recuperation or transformation of velocity to temperature must be known well enough to guarantee an accurate end result.
That site test supplied important data for the present concern. The paper will give the measuring arrangement and the temperature differences measured along with the water flow velocities. These measurements are precious to the future simulation and to understand the mechanism of the velocity effect upon the temperature. Possible conclusions will be drawn.
Case Study:
This paper will present a technical review and discussion of experiences gained during recent flow measurement tests at Dr. Gabriel Terra Power Station, using the current meter method.
The Dr. Gabriel Terra Power Station is located on the Rio Negro, in the centre of Uruguay, and originally entered into full service in 1948, with the four Kaplan generating sets providing a rated power output of 128 MW. Recently these units were fully refurbished, including replacement of the original Kaplan runners (each 32MW and 6-bladed) with new runners (39MW and 5-bladed). This refurbishment has resulted in a maximum increase in power output of 7 MW (22%) per unit, and an average increase of 4% in turbine efficiency. This project is managed by AdministratiÛn Nacional de Usinas y Trasmisioncs ElÈctricas (UTE) of Uruguay, with specialist advice and site supervision services being provided by Merz and McLellan Consulting Engineers of the United Kingdom.
The intakes and penstocks of this scheme are of relatively short length as defined by IEC41, and some degree of non-uniformity of the flowfield at spiral inlet has as a result been suspected. The shape of velocity profile at spiral inlet for one of these turbines was originally measured using 25 current meters arranged on a 4-arm frame, and these results indicated slight skewness of the profile towards the lower held of the penstock. In addition, the penstocks are constructed from riveted plate, and the localized effects of the rivets and overlapping plates are believed to distort the velocity profile near the wall, making it difficult to measure flow precisely using peripheral current meters.
In order to more accurately define this velocity profile and subsequently derive the value of absolute turbine hydraulic efficiency, a specially customized measuring layout was developed, consisting of 60 current meters arranged on an 8-arm frame. The paper will describe the measuring process used and results obtained.
Some of the practical difficulties encountered during the tests are also described, and the paper goes on the example how these were acceptable minimized by good planning in order to undertake performance tests within given time scales.
The following topics are also likely to be covered in the article:
Results from the efficiency tests will be summarized.
The winter - Kennedy piezometers (and their counterpart - the Joseph Peck piezometers) have been widely used for relative flow measurement since their development in the 1920's. Historically, water manometers have been used to measure the pressure differential between a pair of such piezometers, one located on the inner radius of the spiral or semi - spiral case and the other on the same plane, but the outer radius. In recent years, with the advent of newer efficiency testing and recording equipment, the water manometer has been replaced with pressure transducers. The use of either water manometer or pressure transducers for measuring the pressure differential can produce identical relative efficiency test results as long as each test if referenced only to itself and not compared to other tests on the same unit. However, there have been numerous instances of such attempts to compare recent tests utilizing pressure transducers to historical tests with water manometers, which have lead to the conclusion that there has been no change in efficiency over the years or even that the efficiency has increased. This comparison error has now been traced to the readings of pressure transducers, which although precisely calibrated in a laboratory, were not calibrated to the same specific weight of water as existed in the spiral case at the time of the test.
In a related matter, the Winter - Kennedys are often "calibrated" by equating the square root of the pressure differential at peak relative efficiency to the absolute flow rate at peak absolute efficiency, such as that predicted from a model test. However, the volumetric flow rate coefficient derived by such a comparison is not precisely correct unless the rations of the specific weights of water and also of the differences in power at the points of peak efficiency are taken into account.
This paper will present the methods by which the Winter - Kennedys and other relative flow differential piezometric systems can be correctly utilized with the newer electronic test recording equipment and can still be calibrated to provide correct historical comparisons to other tests.
The measurement of temperature at the outlet of the draft tube for thermodynamic measurements is often carried out by only one temperature probe. Measurements carried out at a Norwegian power plant shows that both the temperature and velocity varies over the outlet area of the draft tube. When calculating the hydraulic efficiency this has to be taken into consideration. This is presented in a paper, "Energy distribution at the outlet of the draft tube," at the IGHEM-conference in 1996 by Ole G. Dahlaug and Hermod Brekke.
The uncertainty at the measurement at the outlet of the draft tube is often neglected, and therefore only the uncertainty of the temperature probe itself is taken into consideration when evaluating the overall uncertainty of the thermodynamic efficiency measurement. The uncertainty specified by IEC 41 for temperature measurements at the outlet of the draft tube is ± 0.6% when using only one temperature probe, and measurements carried out at two Norwegian Francis turbines confirms this. This uncertainty can, for high head turbines, be the dominating factor in the overall uncertainty. But how can we calculate the uncertainty of measurements carried out by the collector method or by more than one temperature probe at the outlet of the draft tube? Literature written by Ernest O. Doebelin and Arne Kjolle, and the standards such as ISO-5168 treat the general and specific uncertainty. This paper will discuss how to include and use the different equations for the uncertainty on the measurement taken at the outlet of the draft tube. The uncertainty will also be evaluated for different turbine heads.
Using ACCUSONIC 7510P penstock protection system in two ENEL S.p.A. power plants the acoustic flowmeter behavior was tested in presence of velocity distribution disturbances. IEC 91 standard, asking for a straight length of upstream conduit before the measuring section, are not always observed in protection system because the sections are in positions not designed for this installation, due to the whole penstock protection.
In the installations presented in this paper the measuring sections are positioned in a correct mode in only one of the two measuring sections while the second one is positioned near disturbances ad valves or curves.
In the first of the two power plants tested, the upstream measuring section is positioned according to IEC standard and downstream section is positioned after a 45 curve. In the second power plant there are three penstocks that feed in parallel the hydraulic machinery with various kind disturbances as butterfly valves or connections between the penstocks. The installed acoustic flowmeters are 4 or 2 paths type.
In the paper the differences between upstream and downstream flows with all the paths well running are analyzed and the performances of the flowmeter are tested against anomalies presented in one or more acoustic paths with guaranteed constant discharge. Using the recorded data with various flows a new weight for each acoustic flow was calculated with a reduction of the difference between flows at the beginning and the end of the penstock.
Without doubt the hydraulic parameter which is hardest to measure is flow. The various methods traditionally used to implement this measurement can be both complex and costly. The method using clamp-on ultrasonic flowmeters (the probes are in contact with the outer walls of the penstock) has the advantage of not creating disturbances and of being easy to install. Nevertheless the flow measurement uncertainties announced are around +2 to 3%. We wanted to check these levels of uncertainty by making several comparisons using on site tests and laboratory qualifications. This paper reports on the results obtained, proposes methods for improving measurement accuracy and gives examples for the application of this type of flowmeters.
Our team has acquired Sea-Bird oceanographic temperature probes in order to take efficiency measurements using the thermodynamic method. This paper describes the structures created to use these probes (energy sensor, downstream probe support). It also presents a comparison between the calibration performed by the probe supplier and that carried out in a national metrology laboratory. Finally, two comparative efficiency measurements (Sea-Bird probes/PT100 probes) are presented: One concerns a Pelton turbine with a 770 m head, and the other Francis turbine with a 120 m head. The slight differences between the two methods are fully registered in the measurement uncertainties.
The third edition, dated 1991, of the IEC 41 International Standard does not guarantee uncertainty values on efficiency measurement of a hydraulic machine by using the thermodynamic method. The previous version of the standard guaranteed a maximum uncertainty equal to +1.5% on efficiency measurement for fixed measurement conditions. This value of +1.5% is commonly used on the reception of new runners for hydraulic machines in order to determine the scrap criterion and to calculate and sanctions to be applied to manufacturers.
Based on the recommendations of the IEC 41 and X 07-020 standards (French uncertainty calculation standard), we have used on site test measurements to perform five calculations which highlighted the various values affecting accuracy of flow and efficiency measurement of a hydraulic machine using the thermodynamic method. Efficiency uncertainty decreases as head increases. A statistical law translating this development has been determined. The results of our study have been compared with those obtained by two Japanese colleagues from the Central Research Institute of Electric Power Industry (CRIEPI). Convergence of the results of both studies shows that the level of uncertainty normally accepted, in particular for preparing specifications on the reception of new runners, i.e. +1.5%, is amply overestimated, especially for heads far in excess of 100 meters.
This paper presents field data and laboratory data supporting computational fluid dynamic (CFD) modeling of the effect of acoustic transducer protrusion and conduit on water. Recent tests at a plant in southern California, where relatively good hydraulic conditions exist, indicated that the discharge measurements using two different types of transducers with differing protrusions were reading differently. Initially, the discharge readings based on internal mounted acoustic transducers indicated that the turbine was not performing as required. A second test was performed using "feedthru" mounted transducers located upstream which indicated a lower discharge and greater turbine efficiency. Further investigations using CFD and data obtained at Alden Research Laboratory indicate that the conduit used to route the signal cables away from the internally mounted transducers were causing the discharge measurement to read high. Based on these investigations, new guidelines for installing internally mounted transducers have been adopted.
As part of this investigation additional analysis using CFD and lab tests have confirmed that the effect of transducer protrusion of feedthru mounted type transducers is negligible. This effect was analyze using CFD and validated in the lab. The results of the lab tests indicated that the velocity based on feedthru type types of transducers tend to be slightly negative. The authors will present these findings and present the newly adapted guidelines.
A performance evaluation was conducted to determine the accuracy of 2 acoustic flowmeters installed at Hoover, Davis, and Parker Dams on the lower Colorado River. Field surveys and physical model studies were used to evaluate and enhance the performance of the chordal-path acoustic velocity meters. A hydraulic model and a laser doppler anemometer were used to determine velocity distributions for nonstandard flowmeter installations at Parker and Davis Dams. These velocity distributions were then used to as input to the integration technique used by a chordal-path flowmeter. The results and conclusions of this performance evaluation will be summarized in this paper.
Due to practical and economical limitations the maximal number of acoustic axial planes in typical installations very seldom exceeds the number of four. Especially when the flow profile in a conduit is distorted, this number is to small to deliver the information needed for an error-free integration of the discharge.
In this paper first the mathematical underlyings of the Gauss-Jacobi integration method and of the OWICS method from the ETHZ, which is derived from the Gauss-Jacobi method, will be discussed. Based on distorted flow profiles, which are described by mathematical functions, the integration error will be explained (see figure). To calculate and to predict integration errors, an interpolation tool for scattered data based on triangulation was developped. With this tool, we calculated the integration errors for numerous velocity data taken from field measurements, CFD calculations and analytical velocity distributions. Special emphasis was put on flow fields behind single and double bends. The results show that the integration error very seldom exceeds the range of ±1% and that it can be reduced by selecting the appropriate mounting angle of the ADM for the predicted flow conditions.
Acoustic discharge measurements (ADM) were performed in the tailwater channel of the "Etzelwerk"-powerstation (Switzerland) being equipped with 7 pelton tubines and 2 multistage pumps. Goal of these measurements was the calibration of permanently installed Venturi type discharge meters and the safety devices in the penstock.
An 8 path acoustic flow meter was installed in the tailwater channel, close to its exit into the lake of Z¸rich, 600 m downstream of the powerplant. Simultaneously, the free surface at the measuring section was also measured acoustically with 2 meters.
During the measurement, waves were traveling up- and downstream in the tailwater channel with a period of about 7 minutes. These waves arose from the varying load conditions which had to be adjusted for the individual measuring points and induced low frequency, high amplitude oscillations of discharge at the measuring section. These fluctuations were only lightly damped as displayed for one exemplary operating point in Figure 1.
Because of the high number of operating points to be measured (80), long waiting times between the measuring points of one hour, which would have been needed to achieve steady discharge conditions, were not acceptable.
Since these waves and the associated discharge fluctuations in the tailwater channel did not affect the turbine or pump operation, a special procedure could be developed to determine the mean discharge QADM.
The individual measuring points of discharge were sampled nominally ata frequency of about 0.5 Hz but at spurious instants in time missing data points occurred due to disturbances in the flow. Therefore, in a first step of the procedure to determine the mean flow rate the raw data had to be interpolated to equidistant points in the time domain. This procedure provided additionally a low pass filtering of the data in the ensemble.
In a next step a least square fit was employed the identify the coefficients of a damped Fourier series.
A maximum number of k= 4 in this series showed to be sufficient for an accurate determination of QADM, the discharge being of interest.
Detailed analysis of the errors contributing to the uncertainty of the acoustic discharge measurements was performed and it could be shown that the overall uncertainty will - in spite of the high amplitude fluctuations - not exceed ±1.3 %.
When continuous monitoring of turbine discharge is desirable or required (predictive maintenance, operation of a cascade of power plants,.), acoustic or Winter-Kennedy type of flowmeters are the available options. Even is still not accepted by the IEC for the absolute discharge measurements, the acoustic flowmeter is the better option. It is more accurate, does not rely on the validity of the model - prototype relationship and is ore reliable, since it does not contain any embedded piping, which can break or get clogged. In the case of low head units with short intakes the choice is limited to the Winter-Kennedy type of flowmeter. If higher accuracy is required, it can be calibrated using any suitable Code-accepted method.
Unless incorporated during the construction, the installation of both types of flowmeter may result in difficult,, expensive and not always justifiable. Looking for a reasonable, inexpensive solution we came across the fact that for most of the existing turbines the discharge was determined in the model tests for the full range of net heads and guide vane openings. The theoretical relationship between the prototype discharge and the guide vane opening is therefore available for any normal net head. Tests were conducted with several identical, large Francis turbines, three of them equipped with the acoustic flowmeters. The paper explores the feasibility of continuous discharge monitoring on units not equipped with any type of flowmeter. The prototype test results are analyzed, the validity of the corresponding relationship determined on the model is examined and solutions leading to improved accuracy and repeatability are outlines.
The paper deals with problems caused by uneven temperature and velocity distribution at the outlet.
For Francis turbines cross flow in runner blades and draft tube swirl which are vectors, may create a swirl normal to the axis of the draft tube when changing direction in the draft tube bend. At part load, the instability of the rotating core at the center may create disturbance of both flow and temperature. For Pelton turbines, the air bubbles create a temperature difference and from experiments, high temperatures may occur near the foam on the surface, leading to uncertainty of the result. Concentrated high or low temperature zones are often found in the outlet of Pelton turbines, depending on the geometry of turbine housing. Furthermore the velocity profile often has a skewed character which indicates losses at runner outlet and in the draft tube. In such cases, a manifold collector of water for one temperature gives an error.
In this paper the problems of variation in the outlet energy of the turbines are discussed. Examples of analysis of losses referring to a paper presented at GPMT meeting in Salzuburg by the author, is compared with the analysis of turbine flow and the latest "multi-point" temperatures presented by Dr. Ing. Ole Dahlhaug et al.
The paper presents an evaluation of different losses in a turbine in and invite for a discussion on the possibility to recognize the losses by the energy field at the draft tube outlet.