238

Hydraulic turbine efficiency

J.L. Gordon

Abstract: A set of empirical equations has been developed which defines the peak efficiency and shape of the effi-
ciency curve for hydraulic turbines as a function of the commissioning date for the unit, rated head, rated flow, runner
speed, and runner throat or impulse turbine jet diameter. The equations are based on an analysis of peak efficiency data
from 56 Francis, 33 axial-flow, and eight impulse runners dating from 1908 to the present, with runner diameters rang-
ing from just under 0.6 m to almost 9.5 m. The metric specific speeds (nq) ranged from 5.3 to 294. The root mean
square error of the calculated peak efficiency for Francis and axial-flow runners was found to be 0.65%. The shape of
the efficiency curves was derived from eight Francis, five Kaplan, three propeller, and four impulse turbines. Charts
showing the relationship between calculated and actual efficiency curves for these 20 runners are provided. A good
match between calculated and measured or guaranteed efficiency was obtained. The equations were also used to deter-
mine the relative increase in peak efficiency for new reaction runners installed in existing casings at 22 powerplants,
with a root mean square accuracy of 1.0%. The equations can be used to (i) develop efficiency curves for new and old
runners; (ii) compare the energy output of alternative types of turbines, where this choice is available; and (iii) calcu-
late the approximate incremental energy benefit from installing a new runner in an existing reaction turbine casing, or
onto the shaft of an impulse unit.

Key words: hydraulic turbines, turbine renovation, turbine efficiency.

Resumé : Un ensemble d’équations empiriques a été développé pour définir le rendement de pointe et la forme de la
courbe de rendement pour turbines hydrauliques, et ce, en fonction de la date de mise en service de l’unité, la chute
spécifiée, le débit spécifié, la vitesse de la roue d’eau, et la gorge de la roue d’eau ou le diamètre du jet d’une turbine
à impulsion. Les équations sont basées sur l’analyse de données de rendement de pointe provenant de 56 turbines Fran-
cis, 33 turbines de type axial et 8 turbines à impulsion datant de 1908 jusqu’à aujourd’hui, avec des diamètres de roue
de l’ordre d’un peu moins de 0,6 m jusqu’à presque 9,5 m. Les vitesses spécifiques métriques (nq) vont de 5,3 à 294.
La racine de la moyenne des erreurs au carré du rendement de pointe calculé pour les turbines Francis et de type axial
a été trouvée comme étant 0,65%. La forme des courbes de rendement a été dérivée à partir de huit turbines Francis,
cinq turbines Kaplan, trois turbines à hélices et quatre turbines à impulsion. Des graphiques montrant la relation entre
les courbes de rendement calculées et réelles pour ces 20 turbines sont fournis. Une bonne correspondance entre le ren-
dement calculé et celui observé ou garanti a été obtenue. Les équations ont aussi été utilisés pour déterminer
l’augmentation relative du rendement de pointe pour de nouvelles turbines à reaction installées dans des logements
existants à 22 centrales hydroélectriques, et ce avec une racine de la moyenne de la précision au carré de 1,0%. Les
équations peuvent être utilisées pour (i) développer des courbes de rendement pour de nouvelles et vieilles turbines;
(ii) comparer la production d’énergie de différents types de turbines, lorsque ce choix est possible; et (iii) calculer une
approximation du gain additionnel en énergie provenant de l’installation d’une nouvelle turbine dans un logement exis-
tant de turbine à réaction ou sur l’axe d’une turbine à impulsion.

Mots clés : turbines hydrauliques, rénovation de turbines, rendement de turbines.

[Tradiut par la Rédaction] 253

Introduction Gordon the full operating range of a turbine. Where efficiency is dis-
cussed, the data provided are insufficient to accurately deter-
Hydro turbine efficiency has not been discussed in detail mine the shape of the efficiency curve. To overcome this
in any paper known to the author, and there are no empirical deficiency, the author has developed a series of empirical
methods available to calculate the expected efficiency over equations for calculation of turbine runner efficiency. Ac-
count is taken of the increase in efficiency of newer designs,
and of deterioration since commissioning. Thus estimates
Received April 4, 2000. Revised manuscript accepted October
10, 2000. Published on the NRC Research Press Web site at
can be made of performance of old runners, where there are
[Link] on March 23, 2001. no data on efficiency.
This methodology is primarily intended as an aid in esti-
J.L. Gordon. Hydropower Consultant, 102 St. John’s mating new runner performance at the feasibility study stage
Boulevard, Pointe Claire, QC H9S 4Z1, Canada
(e-mail: jim-gordon@[Link]).
and for estimating old runner performance where it is
impractical to undertake efficiency tests or where commis-
Written discussion of this article is welcomed and will be sioning test records are unavailable. For old runners it is as-
received by the Editor until August 31, 2001. sumed that all similar turbines of a defined age have about

Can. J. Civ. Eng. 28: 238–253 (2001) DOI: 10.1139/cjce-28-2-238 © 2001 NRC Canada
Gordon 239

the same performance, as would be expected in a competi- (1) Zero efficiency is shown at a flow of about 8% of
tive market. For example, in 1978, the author found that the rated flow. This point is called the “speed-no-load” (SNL)
three top bids for a 17 MW turbine all had peak efficiencies flow, the point at which the turbine runner is rotating at syn-
within 0.25%, and the shape of the efficiency curves were chronous speed and ready to be placed on line. Its position
within 1% over the operating range. The analysis develops can vary from a low of less than 1% flow in a high-head im-
both peak efficiency equations and efficiency curves for sev- pulse turbine to as much as 30% of flow or more in a very
eral types of runners. In addition, comparison is made be- low-head propeller turbine.
tween calculated and actual efficiency gains for new runners (2) Peak efficiency occurs at a flow of about 80% of rated
installed in existing units and briefly discusses the effect of flow. This position also varies according to the type of run-
changing head. ner. It is usually at about 65–80% flow in both impulse and
Kaplan runners, varies between 80 and 95% flow in a Fran-
cis runner, and is between 94 and 100% flow in a propeller
runner.
Methodology (3) The value of the peak efficiency also varies, from a
Finding data on turbine efficiency is difficult. Manufac- low of about 80% in low-head old reaction runners and old
turers regard such information as proprietary and only pub- impulse runners to a high of almost 96% in large modern
lish efficiency data as relative, when comparing new with Francis runners.
old runners, with a peak relative efficiency for the old runner (4) The efficiency at full load flow is about 4% lower than
set at unity. At the feasibility study stage, requests for effi- the peak efficiency. This drop also varies between about 1
ciency information often result in a “generic” efficiency and 5%, being less in modern runners.
curve, which only approximates the actual peak and shape of (5) The shape of the efficiency curve varies considerably,
the efficiency curve of the prototype runner installed at the being flattest for impulse and Kaplan runners and almost
development. In such situations, manufacturers are reluctant straight between speed no load flow and peak efficiency
to divulge data, since it could compromise a competitive ad- flow in an old, low-head propeller runner. For a Francis run-
vantage. Even after the project is completed, efficiency data ner the shape lies between these two extremes, being close
are rarely published and, if published, are usually in the rela- to a propeller runner in a low-head Francis turbine, gradu-
tive form with a peak at 1.0. In the search for efficiency in- ally approaching the shape of the Kaplan runner efficiency
formation, the author had to make extensive use of data in curve in a high-head Francis turbine.
his own personal files, old textbooks, papers in International From a mathematical standpoint, the shape of the effi-
Water Power and Dam Construction, papers in the transac- ciency curve can be expressed as a parabola with the apex at
tions of the Canadian Electricity Association (CEA), and the peak efficiency. The efficiency at other flows is then
files from several utilities. In some cases, partial data were equal to the peak efficiency minus a proportion of the peak
found in one publication, which were supplemented with efficiency (Depeak), which is a function of the difference be-
data from another source. Hence, the source of peak effi- tween peak efficiency flow and the considered flow (DQ).
ciency data in Tables 1–3 is not provided, since it would be Therefore to define efficiency for any flow, the equation
too extensive and confusing. However, important references should have two parts, one part to define the value of the
are cited where necessary. peak efficiency, and a second part to define the change in ef-
To facilitate development of efficiency equations for old ficiency with change in flow. The approach was to develop
runners, where data are in U.S. units, the specific speed was the form of a preliminary set of equations for the peak effi-
converted into metric specific speed (nq) calculated with an ciency, test the equations on a few turbines, and then gradu-
assumed turbine full load efficiency 3% below the peak effi- ally refine the equations, adding terms as necessary, using
ciency. The error inherent in this assumption is negligible, as Excel spreadsheets to optimize the terms. The same ap-
will be apparent by using slightly different specific speeds in proach was used for the efficiency shape equations, but opti-
the equations. mization was not possible with so few examples in the
Nearly all published turbine efficiency data are provided database; instead, refinement continued until the difference
as a curve of efficiency plotted against power. In this paper, between calculated and actual efficiency was within ±1.5%.
efficiency is derived as a function of flow, since flow, head, The equations were found to be complex, with several coef-
and efficiency are required to calculate power. However, in ficients and parameters raised to exponents.
the figures where the calculated efficiency is compared with
published data, efficiency is plotted against horsepower or Peak efficiency equation for reaction
kilowatts or the ratio to rated power, if this was the pub-
runners
lished data format. Also, efficiency, as determined by the
equations in this paper, includes the draft tube exit loss, as Peak efficiency was found to be a function of the commis-
defined by the American Society of Mechanical Engineers sioning date for the runner, type of runner, specific speed,
code (ASME 1992). and runner size. Hydraulic turbines were first used to gener-
A typical efficiency curve for a hydro turbine is shown in ate electric power over 100 years ago, using a turbine design
Fig. 1, where the rated flow has a value of unity. The proper- developed by James Francis in 1849. Over the past century,
ties of the curve depend on the type of runner, head on the turbine peak efficiency has increased from under 80% to
turbine, size, rotational speed, and year of commissioning. over 95% in very large Francis turbines. Pfafflin (1987) has
Note the following: shown that peak efficiency of Francis runners has increased

© 2001 NRC Canada

240 Can. J. Civ. Eng. Vol. 28, 2001

Table 1. Comparison of calculated and actual turbine efficiencies for Francis turbines.
Efficiency
Specific speed Runner
Project namea Year builtb nq diameter (m) Actual Calculated Difference (%)
Adam Beck 1922 39.6 2.874 0.916 0.904 –1.2
Case No. 6, old 1925 73.5 2.838 0.920 0.907 –1.3
Martin 1925 65.8 3.453 0.905 0.912 +0.7
Case No. 3, old 1927 69.6 0.841 0.880 0.886 +0.6
Figure 8 1928 66.3 3.222 0.924 0.915 –0.9
Morris 1928 47.5 0.914 0.910 0.892 –1.8
Masson, old 1932 58.8 2.896 0.925 0.920 –0.5
Beauharnois, old 1932 107.5 5.258 0.900 0.911 +1.0
Figure 2/48 1932 105.0 0.600 0.860 0.863 +0.3
Hoover 1933 35.1 3.353 0.918 0.922 +0.4
Case 7, old 1934 62.9 4.200 0.926 0.927 +0.1
Minden 1936 99.7 1.332 0.895 0.894 +0.1
Case No. 5, old 1937 82.1 3.539 0.925 0.925 –0.2
Andrews No. 2 1941 53.0 1.777 0.926 0.923 –0.3
Shasta 1944 40.5 4.343 0.932 0.937 +0.5
Flatiron 1948 23.5 1.411 0.915 0.920 +0.5
High Falls No. 3 1949 69.3 1.777 0.925 0.928 +0.3
Case No. 2, old (H) 1950 74.9 1.500 0.920 0.924 +0.4
Case No. 1, old (H) 1951 130.5 1.561 0.880 0.880 0.0
Smelter 1957 60.8 2.350 0.933 0.938 +0.5
Bratsk 1962 86.1 5.500 0.930 0.943 +1.3
Songa 1964 32.9 2.554 0.947 0.940 –0.7
Grand Coulee No. 3 1965 70.7 9.479 0.950 0.954 +0.4
Murray No. 1 1967 23.3 1.862 0.930 0.934 +0.4
Rana 1968 21.3 1.920 0.937 0.934 –0.3
Churchill Falls 1970 43.9 4.254 0.950 0.950 0.0
Bighorn 1972 63.7 3.252 0.941 0.947 +0.6
Krasnoyarsk 1974 88.0 7.500 0.940 0.949 +0.9
Dadin Kowa 1978 100.8 2.975 0.938 0.931 –0.7
La Grande No. 2 1979 54.8 5.599 0.958 0.954 –0.4
Itiapu 1981 65.4 8.099 0.960 0.956 –0.4
Song-Loulou 1981 99.9 4.310 0.935 0.937 +0.2
Case No. 6, new 1983 1963.7 75.5 2.838 0.940 0.940 0.0
Masson No. 4, new 1983 1966 62.2 2.896 0.944 0.944 0.0
Norway No. 1 1982 110.6 1.600 0.932 0.914 –1.8
Andekaleka 1983 42.5 1.588 0.940 0.940 0.0
La Grande No. 4 1984 60.9 5.547 0.957 0.954 –0.3
Case No. 3, new 1985 1965.7 72.5 0.841 0.930 0.926 –0.4
Case No. 7, new 1985 1968 76.5 3.935 0.949 0.945 –0.4
Case No. 5, new 1985 1969 85.6 3.539 0.940 0.940 0.0
Case No. 2, new 1985 (H) 1973.3 80.0 1.500 0.930 0.933 +0.3
Case No. 1, new 1985 (H) 1973.7 127.1 1.561 0.900 0.894 –0.6
Saucell-Huebra 1985 90.9 5.212 0.947 0.944 –0.3
EPRI 1989 66.1 1.128 0.919 0.936 +1.7
Rattling Brook, new 1985 1976.3 53.4 1.067 0.936 0.935 +0.1
Portal 1987 57.9 1.600 0.947 0.941 –0.6
La Grande No. 4 1988 60.6 5.547 0.957 0.954 –0.3
La Grande No. 2A 1989 60.9 5.233 0.959 0.953 –0.6
Kootenay, new 1994 1987.7 69.6 4.999 0.948 0.952 +0.4
Big Eddy, new 1994 1969.3 119.3 2.185 0.895 0.908 +1.3
Pit No. 1 1995 52.3 1.969 0.940 0.944 +0.4
Spilje, new 1996 1987 63.6 1.887 0.941 0.942 +0.1
Beauharnois, new 1996 1974.7 102.0 5.385 0.937 0.938 +0.1
Gocekaya 1996 69.6 1.152 0.945 0.935 –1.0
Macagua, Nos. 7, 8 1997 100.3 7.657 0.947 0.944 –0.3
Rose Blanche (H) 1998 45.0 0.663 0.925 0.930 +0.5
Note: Total number of runners = 56.
a
H in parentheses indicates horizontal-axis unit with shaft through draft tube.
b
For new runners in old casings, one third of the existing casing age was deducted from the new runner installation date.

© 2001 NRC Canada

Gordon 241

Table 2. Comparison of calculated and actual turbine efficiencies for axial-flow turbines.
Efficiency
Specific speed Runner
Project name Year built nq diameter (m) Actual Calculated Difference (%)
Victor Kaplan 1919 230 0.60 0.810 0.818 +0.8
Green Island (propeller) 1922 193 3.96 0.880 0.888 +0.8
Englesson 1932 179 0.60 0.855 0.858 +0.3
Wheeler (propeller) 1936 184 6.71 0.925 0.916 –0.9
Guntersville 1939 183 6.73 0.929 0.919 –1.0
Austin 1939 150 2.59 0.920 0.903 –1.7
Chicamauga 1940 216 6.71 0.915 0.915 –0.0
Santee Cooper 1942 153 5.33 0.915 0.919 +0.4
Latschau 1943 120 1.90 0.890 0.897 +0.7
Gota Alv 1946 196 5.80 0.925 0.922 –0.3
Deep Brook 1950 182 2.46 0.912 0.913 +0.1
Nimbus 1952 178 3.28 0.920 0.920 0.0
Scott 1952 163 2.65 0.913 0.917 +0.4
McNary 1954 159 7.11 0.942 0.933 –0.9
Grand Rapids 1965 139 6.45 0.940 0.938 –0.2
Kainji No. 7 1967 161 6.45 0.932 0.940 +0.8
Kainji No. 11 1976 133 6.50 0.946 0.942 –0.4
Kainji No. 5 (propeller) 1977 136 6.80 0.940 0.943 +0.3
Tube No. 1 1980 215 1.40 0.928 0.918 –1.0
Old Town (tube) 1980 244 2.20 0.920 0.915 –0.5
Jenpeg (bulb) 1980 294 7.50 0.917 0.913 –0.4
Jebba (propeller) 1982 150 7.09 0.950 0.947 –0.3
Norway No. 2 1982 203 1.60 0.915 0.924 +0.9
Nipawin 1985 142 6.30 0.952 0.945 –0.7
Steephill 1988 103 2.65 0.925 0.928 +0.3
EPRI 1989 159 4.57 0.937 0.943 +0.6
Limestone (propeller) 1989 167 7.94 0.945 0.949 +0.4
Shuikou 1991 95 8.00 0.949 0.941 –0.8
Latschau 1993 143 1.90 0.923 0.933 +0.9
Yacireta (propeller) 1995 189 9.50 0.953 0.950 –0.3
Kalagalaa 1997 162 4.73 0.938 0.944 +0.6
Merowea 1998 153 6.29 0.947 0.946 –0.1
High Falls 2001 120 2.85 0.937 0.934 –0.3
Note: Total number of runners = 33.
a
Feasibility study.

Table 3. Comparison of calculated and actual turbine efficiencies for impulse turbines.
Efficiency
Jet specific Jet diameter
Project namea Year built speed nq (m) Actual Calculated Difference (%)
Kern, old 1908 5.3 0.186 0.820 0.822 +0.2
Turgo No. 1 1919 14.6 0.118 0.820 0.814 –0.6
Tiger Creek 1931 5.7 0.261 0.870 0.871 +0.1
Turgo No. 2 1936 16.1 0.133 0.850 0.854 +0.4
Kern, new 1949 6.3 0.186 0.900 0.890 –1.0
Turgo No. 3 1965 19.7 0.162 0.890 0.893 +0.3
Harca 1968 7.0 0.130 0.902 0.902 +0.1
Cat Arm (V)a 1985 7.0 0.226 0.917 0.918 +0.1
Note: Total number of runners = 8.
a
V in parentheses indicates vertical unit. All other units have a horizontal shaft.

steadily from the beginning of this century to about 1960, stage, with a variety of blade designs, some known as
and thereafter the increase has been slower. Prior to about Nagler wheels and others as Moody wheels where the axis
1920, propeller turbines were very much in the development of the blade was at an angle to the shaft. All had steeply

© 2001 NRC Canada

242 Can. J. Civ. Eng. Vol. 28, 2001

Fig. 1. Typical efficiency curve for a hydraulic turbine plotted as Fig. 2. Increase in peak hydraulic model turbine efficiency over
a function of flow. the past 80 years.

Fig. 3. Relationship between specific speed and peak turbine

sloping efficiency curves, which increased in almost direct model efficiency.
proportion to flow. In response to this deficiency, Viktor
Kaplan developed the variable-pitch propeller turbine and
was granted a patent in 1913 (Hackert 1976). A prototype
model was built in 1919, which produced a peak efficiency
of 81% over a wide range of flows, with a maximum output
of 26.3 kW under 3 m head from a runner with a diameter of
0.6 m, rotating at 500 rpm. No further work was undertaken
until 1920, when the model was used as the basis for the
first commercial installation at Lilla Edet in Sweden, com-
missioned in 1925. The efficiency of Kaplan turbines in-
creased very rapidly in the first 20 odd years, and then more
slowly to the current range of 92–95%. These observations
would imply that the increase in efficiency with respect to
time is a function of age raised to an exponent.
The third parameter is the metric specific speed nq, which third part determines the drop in peak efficiency from the
is defined as most efficient specific speed and is a function of the specific
[1] nq = rpm Qrated0.5hrated–0.75 speed. In effect, the first three parts determine the peak hy-
draulic model efficiency. The last, fourth part determines the
where rpm is the turbine synchronous speed in revolutions increase in model efficiency to the prototype size and is a
per minute; Qrated is the turbine flow at rated head and rated function of the runner throat diameter. The equation thus has
load in m3/s; and hrated is the turbine rated head in metres. the following form:
Turbine efficiency is known to be a function of the spe-
cific speed, with both low and high specific speed turbines [2] e peak = A - De year - De specific speed + De size
having lower peak efficiencies than medium speed machines where A is a constant depending on the type of runner;
(ASME 1996). At lower heads (high specific speeds), losses De year is the change in turbine efficiency due to the year the
in the draft tube become increasingly significant, resulting in unit was commissioned; De specific speed is the change in effi-
a decrease in efficiency with a lower head; at high heads ciency due to specific speed; and De size is the change in tur-
(low specific speeds), flow loss through the runner seals in- bine efficiency (Moody formula) due to size. From the
creases, contributing to lower efficiency. Figure 3 shows the analysis of 89 reaction turbine runners, it was found that A
efficiency in a Francis turbine peaking at a specific speed of has a value of 0.9187 for a Francis runner and 0.904 for
60. For an axial-flow turbine (Kaplan or propeller), the peak Kaplan and propeller (axial flow) runners. The difference in
is at a specific speed of 152. The shape of the curve of the the base level is 1.47%, double the 0.75 difference given in
relationship between relative efficiency and specific speed ASME data (ASME 1996).
suggests an equation for a parabola based on a function of The first adjustment to the efficiency takes into account
the difference between the runner specific speed and the the age of the turbine, with an equation of the following
peak efficiency specific speed. form:
Lastly, larger reaction turbines are more efficient than x
smaller turbines due to the lower effect of friction in large æ 1998 - y ö
[3] De year = ç ÷
runners. Hence, an equation for turbine peak efficiency (epeak) è B ø
must also include a function of runner size.
Based on the foregoing, it was reasoned that the peak effi- where y is the year the unit was commissioned, with the lat-
ciency equation should have four parts. The first part fixes est date being 1998. B and x are constants depending on the
the base level of the peak efficiency. The second part ex- type of runner. B was found to have a value of 187 for a
presses the lower peak efficiency of turbines designed in the Francis runner and 252 for an axial-flow runner, and x a
past and is a function of the plant commissioning date. The value of 3 for Francis and 2.03 for axial flow. The age ad-

© 2001 NRC Canada

Gordon 243

Fig. 4. Effect of increased runner diameter on turbine efficiency, exponent. The variation in efficiency with specific speed is
based on a 0.3 m model. shown in Fig. 3.
The third adjustment takes into account the size of the
runner and is based on the step-up formula developed by
Moody (1952). The equation has the following form:

[5] De size = size adjustment

= (1 - A + De year + De nq )(1 - 0.789d -0.2 )

where d is the runner throat diameter in metres. Note that in

eq. [5] the age and specific speed adjustments are always
added, because the Moody adjustment is a function of run-
ner losses. Equation [5] is the standard step-up formula de-
veloped from the laws of friction, with an exponent of 5. It
has the effect of increasing the prototype efficiency as shown
justment has the effect of decreasing the peak efficiency in in Fig. 4.
percentage points as shown in Fig. 2. For Francis runners, a Two spreadsheets were developed for the data, one for
preliminary equation was developed from the Pfafflin (1987) Francis runners and one for axial-flow runners. All coeffi-
data and from other data in several recognized standard texts cients and exponents in the equations were optimized until
such as Barrows (1943) and Creager and Justin (1949). For there was no change in the fourth decimal place of the root
axial-flow runners, a preliminary equation was developed mean square error between calculated and actual peak effi-
from an examination of old Kaplan runner peak efficiencies, ciencies as shown in Tables 1 and 2. To determine whether
starting with data from Victor Kaplan’s original experimen- there is a bias in the error, the error was plotted against tur-
tal runner. For new runners, this age adjustment is omitted. bine commissioning date, specific speed, and runner diame-
For old runners, the age factor is based on the commission- ter. No bias was detected. The sum of the error is 1.4% for
ing date; design date would have been preferable, but it is the Francis runners and –1.1% for the axial-flow runners.
usually unknown. Where there is evidence that the runner Root mean square error is 0.65% for both Francis and axial-
was based on a hydraulic model developed several years flow runners.
prior to the commissioning date, then this date should be
used instead (e.g., in a plant extension, where new runners Peak efficiency equation for impulse
are duplicates of the initial runners). For new runners the runners
date to be used in all time-dependent equations is 1998. It is
recognized that advances are still being made in runner effi- The same methodology was used to develop a series of
ciency, but with peak efficiencies now approaching 96% in equations for impulse runners, but due to the very few exam-
large runners, the advances will be measured as fractions of ples found in the literature, optimization of the factors was
a percent, hence the equations should provide accurate data limited. However, there are two minor differences in the
for many years to come. methodology. First, the specific speed as calculated in eq. [1]
The second adjustment takes into account the specific speed is based on the rated flow divided by the number of jets.
of the turbine runner. The equation has the following form: Second, there is no significant increase in efficiency due to
size (Grein et al. 1988), and this small increase is based on
z the jet diameter, which can be calculated from
éæ n - C ö 2 ù
[4] De nq = specific speedadjustment = ê ç q ÷ ú 0.544Q0.5
êè D ø ú [6] d jet =
ë û h 0.25

where C, D, and z are all constants depending on the type of where djet is the jet diameter in metres, Q is the turbine
runner. The parabolic form of this equation reflects the rated flow in m3/s, and h is the rated head on the turbine in
lower peak efficiencies at both high and low specific speeds metres. The peak efficiency equation is then
and was initially based on the ASME data. The numerator C 2. 5 2
equals 52 for Francis and 162 for axial-flow runners. The æ 1998 - y ö æ 17.5 - n q ö
[7] e peak = A - ç ÷ -ç ÷ + 0.1d 2jet
denominator D equals 292 for Francis and 533 for axial-flow è 240 ø è 185 ø
runners. The constant C indicates peak efficiency at a spe-
cific speed of 52 for Francis runners, which is slightly lower The correlation between this equation and the examples
than the peak at 60 as determined by the ASME. Also, the found in the literature is shown in Table 3. The age decre-
peak at 162 for axial-flow runners is slightly higher than the ment was initially based on Dumbleton (1988) and old effi-
152 peak in the ASME data. The exponent z has a value of ciency data found in manufacturer’s brochures. The specific
1.017 for a Francis runner and 0.979 for an axial-flow run- speed decrement was taken, without change, from the ASME
ner. The exponent with a value of 2 is required to produce a chart (ASME 1997), and the small size increment was based
positive rational number when nq is lower than C. The expo- on the Cat Arm data, assuming the exponent is based on the
nent z was added when optimization indicated that accuracy bucket area, which is a function of the runner diameter. Only
could be increased with a slight variation in the value of the the base level A and the age factors were optimized, since

© 2001 NRC Canada

 244
Table 4. Comparison of calculated and measured peak turbine incremental efficiencies for new runners installed in existing casings.

Specific speed Runner

nq diameter (m) Incremental peak efficiency (%)
Year Year Runner
commissioned uprated Project name typea Old New Old Newe Actual Calculated Difference Reference
1908 1985 Case No. 4 F 29 37 0.79 13.7 12.5 –1.2 Pfafflin 1987
1920 1995 Eglisau F 108 111 3.75 7.0 7.6 +0.6 Sebestyen and Keck 1995
1936 1984 Pickwick A 209 236 7.42 4.2 3.5 –0.7 Wachter 1998
1937 1992 Iller A, Straflo 254 252 2.05 6.5 5.5 –1.0 Sulzer Hydro 1996
1942 1993 Wilson F 85 90 3.82 2.0 3.0 +1.0 Voith Hydro Inc. 1992
1943 1993 Latschau A 120 143 1.90 3.3 5.1 +1.8 Holzer and Schedler 1994
1944 1987 Verbois A 168 171 4.50 4.80 1.6b 3.9 +2.3 Vuillioud and Walther 1998
1944 1994 Midskog A 164 169 6.30 3.8 3.8 0.0 Kvaerner Energy 1997
1948 1982 Aguasabon F 49 51 2.04 1.7 2.8 +1.1 CEA 1990
1951 1992 McMcormick F 83 86 4.23 1.5 2.1 +0.6 Bhan and Bhan 1997
1951 1993 Kananaskis A 156 161 2.74 2.5 3.6 +1.1 McLaughlin 1994
1951 1997 Rosenau A 207 230 4.30 4.70 1.9 2.6 +0.7 Grunder et al. 1997
1955 1986 Bersimis 1 F 34 37 2.55 1.7 2.4 +0.7 CEA 1990.
1957 1998 Losenstein A 178 196 4.50 2.4 2.7 +0.3 Jungbauer et al. 1992.
1958 1989 Orlik A 94 93 4.60 6.4c 2.8 –3.6 Vitvar 1989
1959 1985 Rattling Brook F 44 55 1.07 4.0 2.3 –1.7 CEA 1990
1962 1994 Krokstrommen A 113 115 3.60 3.3 2.8 –1.5 Kvaerner Energy 1997
1965 1985 Hoover F 39 42 3.35 1.0 1.9 +0.9 Wachter 1998
1965 1994 Grand Rapids A 139 137 6.45 1.5 2.0 +0.5 Lambert and Jarriand 1995
1965 1994 Saunders A 137 145 6.40 3.4 2.5 –0.9 Markovich et al. 1991
1965 1996 Bay d’Espoir F 44 45 2.42 3.5 2.0 –1.5 Kepler and Mallam 1997
1969 1996 Spilje F 56 64 1.90 2.00 4.0d 1.8 –2.2 Nestorov and Pavleski 1997
Note: Total number of runners = 44.
a

Can. J. Civ. Eng. Vol. 28, 2001

A, axial; F, Francis.
b
Upgrade for capacity, not efficiency.
c
Old runner had significant cavitation damage.
d
Large increase due to removal of draft tube cross air vent pipes, not required with new runner.
e
Diameter only indicated if changed from original installation.
© 2001 NRC Canada
Gordon 245

there were insufficient data to justify further optimization. and the sum of the error is –1.7%. For a best fit, 1.5% was
Turgo impulse turbines are very rare, with only two or three subtracted from the calculated efficiency of the old runner to
installed in Canada, the first being a 4.5 MW, 237 m head allow for rust and deterioration, but strangely, no such ad-
unit at Revelstoke, British Columbia. No published data, apart justment was required for the Francis runners in Table 1. It
from a manufacturer’s brochure, were found. is suspected that for the Francis runners, the published data
For Pelton units, the base level A has a value of 0.917 for compare new runner efficiencies, whereas the comparative
vertical-axis units and 0.9085 for horizontal-axis, two-jet efficiencies are based on index testing of the deteriorated old
units. The difference is due to the higher distributor losses in runner prior to upgrading and index testing of the new run-
horizontal-axis units. In such units, the pipe supplying the ner.
top jet has to turn upwards through an angle of about 75° This is illustrated by the Orlik case, where the calculated
and then down through an angle of about 120° to impact the efficiency increase, including 1.5% for runner deterioration,
runner at a slightly higher level than the lower jet. The addi- is only 2.5% compared with 6.4% found by before and after
tional head loss around the turns and higher impact level re- index testing. At Orlik, the original 10-blade Kaplan runner
sults in an efficiency loss of about 1.5% for the top jet, was found to suffer from excessive cavitation, and deteriora-
relative to the lower jet, reducing the average efficiency by tion of the efficiency was detected from comparative index
about 0.75% in horizontal-axis units compared with vertical- tests in 1962 and 1966, during which time the relative peak
axis units. However, for single-jet horizontal units, there is efficiency decreased by 1.7%. The new runner has eight
no such loss. For Turgo units, which are more inefficient blades.
than Pelton units, the base level A has a value of 0.900, and Another interesting observation is that in attempting to
the year divisor is 210. verify the equations, the CEA (1990) report data were used
to calculate peak efficiency improvement for runner retrofits.
Efficiency improvement with new reaction The calculated improvement did not agree with the mea-
sured data, which in almost half of the cases recorded either
runners in old casings a peak efficiency decrease or an increase of less than 1.5%
The equations can be used to estimate the probable in- with the new runner. In all these cases, the retrofit was under-
crease in efficiency provided by a new runner, and conse- taken prior to 1985, without the aid of homologous model
quently the probable incremental energy benefits. In reaction tests. After 1985, computational fluid dynamics began to be
units, however, the shape of the bedded parts, and in particu- used extensively to design runners, taking into account the
lar the draft tube, may not be of the correct proportion for a shape of the existing old water passages, and thus vastly im-
modern runner, hence the full benefit of a calculated in- proving the accuracy of runner performance predictions, which
crease in efficiency may not be obtained. On the other hand, illustrates the importance of this new technology (Sabourin et
the old runner is likely to be in a deteriorated condition. If al. 1995).
manufactured from carbon steel, rust will increase roughness Data on only one impulse unit retrofit could be found, for
and reduce efficiency by one or two percentage points Kublis (Gujan and Hauser 1994), where the new runner effi-
(Nichtawitz 1997). Cavitation and erosion will also reduce ciency increased by 4.4% compared with a calculated in-
efficiency. The author has found that a reasonable estimate crease of 6.6%.
of the efficiency of a new runner in an existing casing can be
obtained by allowing for only two thirds of the age effi- Shape of Francis turbine efficiency
ciency increment. To clarify this point, if the original runner
equation
was installed in 1938, for example, and the new runner in
1998, the time difference is 60 years. Subtract one third, or Any set of equations describing the shape of the efficiency
20 years, and use an effective time of 1978 for the new run- equation must take into account three characteristics of
ner installation in all time-dependant equations. An inspec- Francis turbines as follows: (i) the peak efficiency flow rela-
tion of the condition of the old runner is then used to tive to the rated flow changes as the specific speed changes;
estimate the peak efficiency loss, in the order of 1–2% for (ii) the shape of the efficiency curve becomes flatter as the
rusting, or more if there is evidence of serious cavitation or specific speed decreases; and (iii) the speed-no-load flow
erosion. This additional loss is assumed to be proportional to relative to rated flow position decreases as the specific speed
the flow squared in the shape equations. decreases.
Table 1 includes 11 cases where the calculated new Fran- The change in peak efficiency gate opening and the change
cis runner efficiency can be compared with the published in shape of the efficiency curve with respect to the specific
measured or guaranteed efficiency. In all cases, one third of speed have been well documented. Most texts on hydro
the age difference, new to old, was subtracted from the new power have charts showing the different shapes of the effi-
runner installation date. Unfortunately, as mentioned previ- ciency curve for different specific speeds. This effect can be
ously, manufacturers are not publishing absolute data on the approximated by the following equations:
increased efficiency of new runners installed in existing set-
tings. Instead, the relative efficiency is provided. Twenty-two Qpeak
[8] = 1.2n q-0.075
cases were found where the incremental peak efficiency could Qrated
be calculated for comparison with manufacturer’s data. The
results are shown in Table 4. In all calculations, one third of where Qpeak is the turbine flow at peak efficiency and rated
the time difference was subtracted from the commissioning head in m3/s, and Qrated is the turbine flow at rated head and
date for the new runner. The root mean square error is 1.0%, rated load in m3/s.

© 2001 NRC Canada

246 Can. J. Civ. Eng. Vol. 28, 2001

As head decreases, the Francis turbine becomes relatively which excludes the draft tube exit loss in the efficiency. At
more inefficient at part loads, and the SNL gate opening in- Kootenay Canal, one third of the time difference between
creases. Based on data from Lowy (1949), the flow at SNL, new and old turbine commissioning dates was subtracted
as a ratio of the rated flow, can be estimated from from the new turbine commissioning date to obtain the peak
1. 2 efficiency. Except for Holtwood, the Francis efficiency curves
Qsnl æn ö are mostly within the 1.5% margin of error usually accepted
[9] =ç q ÷
Qrated è 570 ø for turbine efficiency. At Holtwood, which has a full Moody
draft tube, the calculated peak efficiency is 3.8% below the
where Qsnl is the turbine flow at SNL and rated head in m3/s. measured efficiency. The Holtwood data are based on Rog-
If the SNL flow is known in an existing turbine, this ers and Moody (1925), and no mention is made of the
should be used instead of the value derived from eq. [9]. Ef- method used to measure prototype performance, which is
ficiency curves for eight Francis turbines were located, and difficult in large low-head powerplants (Doering and Gawne
based on an analysis of these curves, the change in shape of 1998). Even with modern acoustic instruments, accuracy of
the efficiency curve can be reproduced by the following only ±2% is just attainable (Doering and Gawne 2000). It is
equation for flows less than the peak efficiency flow: suspected that model test data were used with an overly gen-
erous allowance for efficiency increase due to size. How-
[10] e q = e peak - De peak ever, the shape of the calculated efficiency curve matches
where e q is the turbine efficiency at flow Q, e peak is the tur- that of the prototype curve right down to SNL.
bine peak efficiency, and De peak is the change in turbine effi- The art of turbine design has advanced over the years,
ciency from peak efficiency: particularly since 1985, with the introduction of computers
to refine runner blade shape and increase accuracy of perfor-
-k k
æ Q ö æ ö mance predictions. Prior to about 1950, it was common to
[11] De peak = e peak çç1 - snl ÷÷ ç1 - Q ÷ include a generous overgate capacity, with the rated flow set
ç Q ÷
è Qpeak ø è peak ø at about 85% of wicket gate stroke. This provided a safety
margin in case the turbine did not produce the rated capacity
where the flow exponent k has a value of at the design flow and some extra flow capacity for low-
[12] k = 3.94 – 0.0195nq head conditions. Of course, this additional flow capacity re-
quired a larger runner, adding to cost. Also, manufacturing
The flow exponent k increases with decreasing specific methods have recently improved with the introduction of
speed, in effect becoming larger at higher heads, producing a fabricated runners, and five-axis milling of the blades, to
flatter efficiency curve. Equation [11] may appear to be produce a precise match to the design shape. As design and
overly complex, but it is only an expression of the form ayx, manufacturing methods improved, overgate capacity was
a parabola with the apex at the peak efficiency and passing gradually reduced and currently is in the region of only
through the zero efficiency point at SNL. Since both the about 3% in large turbines and about 10% in smaller cast
peak efficiency point and SNL are defined as functions of runner turbines. In a cast runner, shrinkage on cooling re-
flow, the value of the coefficient a is fixed as shown in duces runner diameter slightly, hence a more generous over-
eq. [11]. gate capacity is usually required in case shrinkage is more
At flows larger than the peak efficiency flow, to the rated than expected. The extent of overgate capacity in a turbine
flow and over, the decrease in efficiency is expressed by the can be detected with an index test. The relative efficiency
following equation: will drop rapidly as the flow increases to the maximum, often
1. 5 to such an extent that there is no increase, or even a decrease
é æ 1998 - y ö 2 ù æ Q ö in power. The efficiency equations do not take overgate ca-
[13] De peak = êç ÷ + 0.4ú ç - 1÷
çQ ÷ pacity into account, but the loss in efficiency can be calcu-
êë è 100 ø úû è peak ø lated with eq. [13] for higher than rated flows. For example,
the last point calculated for the output at Shasta, at about
The age factor in the coefficient has the effect of increas- 113 000 hp (1 hp = 745.6999 W) in Fig. 7, is at a flow 7.5%
ing the efficiency drop to full load in older runners. The flow over the rated flow.
factor also increases the efficiency drop for runners with the
peak efficiency at a lower proportion of the full load flow.
With eqs. [8]–[13], it is possible to develop curves of effi-
ciency versus either power or flow right down to the SNL Shape of Kaplan turbine efficiency equation
flow. These equations were derived from data for the eight
Francis turbines listed in Table 5. The match between calcu- Equations for an efficiency curve must take into account
lated and actual efficiency is shown for these turbines in two characteristics of Kaplan turbines. The first is the loca-
Figs. 5–13. For Shasta, the efficiency was tested at a head tion of the peak efficiency flow, which varies between about
15% above the rated head of 330 ft (1 ft = 0.3048 m). The 72 and 80% of rated flow, depending on the characteristics
test data were converted to the rated head by reducing the of the runner and the extent of reservoir drawdown. The sec-
flow in proportion to the square root of the head, and the ef- ond is the location of the SNL flow, which ranges between
ficiency was increased using eq. [25]. For La Grande No. 2, about 5 and 12% of rated flow and was derived from the
part of the difference is due to the sonic tests being under- Lowy (1949) data. These characteristics are expressed in the
taken to the International Electrotechnical Commission code, following equations:

© 2001 NRC Canada

Gordon 247

Table 5. Turbine characteristics for shape of efficiency curves.

Runner No. Rated Rated
Turbine Year diameter of head flow Speed Specific
Project name typea built (m) jets (m) (m3/s) (rpm) speed Reference
Holtwood F 1924 4.73 16.5 100.5 94.7 110.4 Rogers and Moody 1925
Martin F 1925 3.45 44.2 88.0 120.0 65.3 Gibson 1958
Shasta F 1944 4.39 100.6 104.5 138.5 40.8 Allis Chalmers 1954
Flatiron F 1948 1.41 321.6 12.8 514.3 24.2 USBR 1976
Bighorn F 1972 3.25 74.7 82.4 180.0 64.3 Thicke and Bakar 1972
Dadin Kowa F 1978b 2.87 29.4 65.7 157.9 101.4 Author’s files
La Grande No. 2 F 1979 5.60 137.2 265.0 133.3 54.2 Mahe and Miron 1991
Kootenay Canal F 1985c 5.00 74.7 204.0 128.6 72.3 Lang and Thomas 1996
Santee Cooper K 1942 5.33 21.0 170.6 120.0 159.6 Allis Chalmers 1954
Nimbus K 1952 3.28 12.7 65.0 150.0 180.3 USBR 1976
Kainji No. 11 K 1975 6.50 38.1 308.2 115.0 132.1 Author’s files
Merowe K 1991d 6.30 35.3 305.0 115.0 139.2 Author’s files
High Falls K 2001 2.85 44.4 56.0 277.0 120.5 Author’s files
Green Island N 1922 3.96 4.0 44.0 80.0 189.0 Barrows 1943
Wheeler P 1936 6.71 14.6 264.5 85.7 186.6 Davis 1952
Jebba P 1982 7.09 29.3 380.0 93.7 145.0 Author’s files
Tiger Creek I 1931 2 362.7 8.78 225.0 5.7 Davis 1952
Turgo No. 3 I 1965 1 228.6 1.34 1000.0 19.7 Gilbert Gilkes and Gordon 1973
Harca I 1968 4 346.3 4.22 500.0 6.4 Gordon 1970
Cat Arm I 1985 6 380.5 20.17 327.3 7.0 Newfoundland and Labrador
Hydro, personal
communication, 1998
a
F, Francis; I, impulse; K, Kaplan; N, Nagler; P, propeller.
b
Powerhouse built, equipment not installed.
c
Retrofit.
d
Feasibility study.

Fig. 5. Comparison between calculated and actual efficiency Fig. 7. Comparison between calculated and actual efficiency
curves for the Holtwood Francis turbine. curves for the Shasta Francis turbine.

Fig. 8. Comparison between calculated and actual efficiency

Fig. 6. Comparison between calculated and actual efficiency curves for the Flatiron Francis turbine.
curves for the Martin Francis turbine.

© 2001 NRC Canada

248 Can. J. Civ. Eng. Vol. 28, 2001

Fig. 9. Comparison between calculated and guaranteed efficiency Fig. 13. Comparison between calculated and actual efficiency
curves for the Bighorn Francis turbine. curves for the Santee Cooper Kaplan turbine.

Fig. 10. Comparison between calculated and guaranteed effi- Fig. 14. Comparison between calculated and guaranteed effi-
ciency curves for the Dadin Kowa Francis turbine. ciency curves for the Nimbus Kaplan turbine.

Fig. 11. Comparison between calculated and sonic-tested effi-

Fig. 15. Comparison between calculated and guaranteed effi-
ciency curves for the La Grande No. 2 Francis turbine.
ciency curves for the Kainji No. 11 Kaplan turbine.

Fig. 12. Comparison between calculated and sonic-tested effi-

ciency curves for the Kootenay Canal Francis turbine.
Fig. 16. Comparison between calculated and guaranteed effi-
ciency curves for the Merowe Kaplan turbine.

Qpeak
[14] = 1.9n q-0.2
Qrated
is also reproduced by eq. [11], but the exponent k has a
value of
0. 5
Qsnl æ nq ö 2
[15] =ç ÷ æ 1998 - y ö
Qrated è 20 000 ø [16] k = 7.2 - ç ÷
è 45 ø

The change in shape of the efficiency curve for a Kaplan The age factor in the exponent is required to take into ac-
unit for flows both above and below the peak efficiency flow count the increased flattening of the efficiency curve over

© 2001 NRC Canada

Gordon 249

Fig. 17. Comparison between calculated and guaranteed effi- Fig. 18. Comparison between calculated and guaranteed effi-
ciency curves for the High Falls Kaplan turbine. ciency curves for a standard small Kaplan turbine.

time as more efficient designs were developed. The Kaplan flow. Since then, the hydraulic design has improved,
efficiency equations were developed from data published on resulting in a slightly more curved efficiency shape and a
five Kaplan runners, all of which are shown in Figs. 14–18. peak efficiency closer to 95% rated flow, as derived in the
Again, a reasonable match was obtained, within the accept- following equation:
able 1.5% error range. The only significant difference is in 2
the lower flow efficiency at Kainji, where runner blade rota- Qpeak æ 1998 - y ö
[17] =ç ÷ + 0.945
tion stops short, thus reducing low-flow efficiency. This is Qrated è 325 ø
due to minimizing the size of the runner hub, to maximize
the flow capacity, and thereby limiting blade servomotor The SNL gate position can be derived from the following
stroke. formula:
However, due to the complexity of a Kaplan design, with 2
the runner blade servomotor in the hub, manufacturers use a Qsnl æ n ö
[18] =ç q ÷
developed design over a wide variation of flows and heads, Qrated è 350 ø
which adds to the difficulty of developing empirical effi-
ciency curves. This is particularly so for small standardized Efficiency at part flows, both above and below the peak
turbines, as illustrated in Fig. 18, which shows the large dif- efficiency flow, can also be derived from eq. [11], and for
ference between calculated efficiency and the guaranteed ef- propellers the exponent k has the following value:
ficiency for a small Kaplan runner of about 5 MW capacity. 2
The difference could be because of an incorrect relationship æ 1998 - y ö
[19] k = 1.78 - ç ÷
between wicket gate opening and runner blade angle or lo- è 92 ø
cating the rated flow well off to one side of the runner hill
chart. Again, the time component in eq. [19] is required to allow
Also, from a theoretical standpoint, a Kaplan efficiency for the increase in efficiency as designs improved. The equa-
shape equation should be made up from two distinct curves, tions were developed from an analysis of three propeller
one for the Kaplan blade in the flattest position, and the runners, and the match between calculated and actual effi-
other describing the locus of the peak efficiency point as the ciency is shown in Figs. 19–21. Propeller turbines have a
blades open with increasing flow. With a database of only rapid drop in efficiency as flows increase above the peak ef-
five runners, it was not possible to derive two curves, and a ficiency flow, hence there is no incentive to provide any
compromise was reached with a single curve. Moreover, it is overgate capacity, which is usually limited to about 3%.
believed that the Kaplan efficiency curve becomes more With only three propeller runners in the database, an effi-
peaked as the specific speed increases, and considerable time ciency match well within the acceptable error was possible,
was spent attempting to include the specific speed in the although the overall confidence in the equation is lower.
equations, but with no success. The database is just too
small and the data are too contradictory to allow for inclu- Shape of impulse turbine efficiency
sion of the specific speed. equation
With Kaplan runners it is possible to provide considerable
overgate capacity, since the drop in efficiency at flows over An impulse turbine consists of a jet of water impinging on
the peak efficiency flow is gradual. Hence an overgate of a series of buckets bolted or cast onto a disk. The concept
15–25% may be used to compensate for lower head, pro- was first envisaged by Leonardo da Vinci in about 1485
vided runner cavitation limits are not exceeded. (Escher Wyss 1981). The split-bucket design was patented
by L.A. Pelton in 1889. Power varies by changing jet diame-
Shape of propeller turbine efficiency ter. Gross efficiency is constant up to the point of peak effi-
equation ciency, with gross efficiency defined as the ratio of energy in
to energy out. Above peak efficiency, the buckets become
About 70 years ago, propeller turbines had an efficiency saturated, and there is a decrease in efficiency, the rate of
curve shape which was almost a straight line when plotted which increases with increasing flow. Up to peak efficiency,
against flow, and the peak efficiency occurred at the rated the shape of the efficiency curve can be derived from this

© 2001 NRC Canada

250 Can. J. Civ. Eng. Vol. 28, 2001

Fig. 19. Comparison between calculated and guaranteed effi- Fig. 21. Comparison between calculated and guaranteed effi-
ciency curves for the Green Island propeller turbine. ciency curves for the Jebba propeller turbine.

Fig. 20. Comparison between calculated and guaranteed effi-

ciency curves for the Wheeler propeller turbine.
flow is the same, no matter how many jets are operating.
The more inefficient Turgo units have a higher SNL flow.
From the foregoing, the following general efficiency equa-
tion can be derived for use up to the peak efficiency:

æ Q Q ö
[21] e q = e peak çç1 - snl + snl ÷÷
è Q jQpj ø

where j is the number of jets in operation on one runner, and

Qpj is the peak efficiency flow of one jet.
The peak efficiency flow per jet was found to be a func-
knowledge. For example, if the SNL flow is at x m3/s, the tion of the number of jets and the age of the unit. The reason
net efficiency (defined as the energy available to produce for this is thought to be the reduction in rotational losses in
electricity divided by energy in) is zero, but the gross effi- more modern generators, which allowed manufacturers to
ciency is much above zero, since energy is being used to decrease the peak efficiency flow ratio to take advantage of
maintain rotation. An increase in flow to 2x m3/s will mean higher efficiency at low flows in a modern unit. The peak ef-
that half the flow is being used to maintain rotation, and the ficiency flow ratio is thus
other half is available to produce power. Since gross effi-
ciency is constant, the net efficiency is 50% times the gross é æ 1998 - y ö ù Qrated
2

efficiency. Losses in the runner, during the conversion of the [22] Qpj = ê 0.63 + ç ÷ ú
êë è 150 ø ú j
jet kinetic energy into torque, are in the region of 6–10% or û
more, depending on the design and age of the runner. For a
runner with a gross efficiency of 90%, the efficiency at a A modern single-jet unit would then have the peak effi-
flow of 2x m3/s is then = 0.5 × 0.9 = 0.45, or 45%. Simi- ciency at a flow of 63% of rated flow, and this ratio would
larly, at a flow of 4x m3/s, the net efficiency is 0.75 × 0.9 = increase to about 70% in a 40 year old unit.
0.675, or 67.5% Theoretically, the peak efficiency of each jet in a multi-jet
Accordingly, all that is needed to derive the shape of the vertical-axis turbine should be about the same, and this is
efficiency curve up to the point of peak efficiency is the SNL the case. However, with all six jets operating in a six-jet tur-
flow and the ratio of peak efficiency flow to rated flow. The bine, all jets contribute to the task of overcoming rotational
SNL flow is a function of the power required to overcome losses. With only one jet operating, the energy required for
friction and windage losses in the generator, which are this task is six times greater, and hence the peak efficiency
higher in older units due to the larger diameters of older with only one jet operating is much lower. This results in the
generators. Based on an examination of the Cat Arm, Harca, following formula for peak jet efficiency in a multi-jet run-
and Tiger Creek efficiency curves, and manufacturer’s Turgo ner:
data, the SNL flow can be derived from the following equa-
tion: æ Q Q ö
[23] e pj = e peak çç1 + snl - snl ÷÷
é æ 1998 - y ö
2ù
è Qrated jQjet ø
[20] Qsnl = ê B + ç ÷ úQrated
êë è C ø ú
û
where Qjet is the rated flow per jet, and j is the number of
where B has a value of 0.0035 for Pelton units and 0.018 for jets in operation.
Turgo units. The divisor C has a value of 700 for Pelton Above the peak jet efficiency point, the buckets become
units and 400 for Turgo units. For a new Pelton unit the SNL saturated and efficiency starts to decrease. The efficiency
flow is at 0.35% of rated flow, and for a 40 year old unit the above the peak efficiency flow can be described by the fol-
SNL flow is at 0.68% of rated flow. Also, note that the SNL lowing formula:

© 2001 NRC Canada

Gordon 251

Fig. 22. Comparison between calculated (thick line) and pub- Fig. 24. Comparison between calculated (thick lines) and guaran-
lished (thin line) efficiency curves for the horizontal-axis, two- teed (thin lines) efficiency curves for the horizontal-axis, four-jet,
jet, two-runner Tiger Creek impulse turbine. two-runner Harca impulse turbine. Guaranteed efficiency curves
shown for two- and four-jet operation.

Fig. 23. Comparison between calculated (thick line) and guaran-

teed (thin line) efficiency curves for the horizontal-axis, single- Fig. 25. Comparison between calculated (thick lines) and guaran-
jet Turgo No. 3 impulse turbine. teed (thin lines) efficiency curves for the six-jet, vertical-axis Cat
Arm impulse turbine. Guaranteed efficiency curves shown for
one-, two-, four-, and six-jet operation.

ì é 1. 5 ü
ï ê æç Q ö÷ ùú ï
[24] e = e peak í1 - Aç1 - ý where De head is the change in turbine efficiency due to
ï êë è jQpj ÷ø ú ï
î û þ change in head from rated head, and h is the net head on the
turbine in metres.
For Francis turbines, however, eq. [25] can only be re-
where A has a value of 0.03 in Pelton units and 0.15 in
garded at preliminary, since the change in efficiency for
Turgo units, Figs. 22–25 show the comparison between cal-
heads other than the rated head is also a function of the spe-
culated (thick lines) and actual efficiencies (thin lines) for
cific speed, with the decrease in efficiency becoming larger
four impulse turbines. The close agreement is due to fitting
as the specific speed increases. Unfortunately, there are in-
the equations to only a few runners. More data are required
sufficient data on this aspect to formulate a more compre-
to determine accuracy.
hensive equation.

Effect of head on efficiency Discussion

The only variable remaining in the efficiency equation is Equations have been developed to determine the effi-
the variation of efficiency with change in head. A propeller ciency of both new and old turbine runners. The peak effi-
runner should not be used where there is a significant ciency equation has been tested against 141 runners in units
change in head. For impulse runners, the change in head rel- dating back to 1908, with runner diameters ranging from 0.6
ative to the rated head is not significant, hence there is no to 9.45 m. For reaction units, the accuracy is mostly within
change in efficiency. For Kaplan and Francis turbines, at ±1.5%, the acceptable error for measurement of efficiency.
constant flow, there is only a minor change in efficiency There are insufficient data to determine accuracy of the im-
with a change in head, as shown by the head–discharge pulse peak efficiency equations. Old reaction units built be-
Figs. 33, 35, 38, and 39 in the United States Bureau of Rec- fore about 1940, and particularly double runner horizontal-
lamation monograph (USBR 1976). Based on these data, the axis Francis units with shafts through the draft tube, may
efficiency at other heads, to about 25% above and below have a lower peak efficiency than that predicted by the for-
rated head, can be estimated from the following equation: mulae, due to high losses in the draft tube. Also, runners
may have been designed at a much earlier date than that in-
2
æ h - hrated ö ferred from the commissioning date. For example, the
[25] De head = -0.5ç ÷ Beauharnois No. 1 powerhouse in Quebec was commissioned
è hrated ø between 1932 and 1948. The No. 2 powerhouse commis-

© 2001 NRC Canada

252 Can. J. Civ. Eng. Vol. 28, 2001

sioned between 1950 and 1953 contains identical runners to nadian Electric Association and the LaSalle Consulting
the first powerplant, hence the earlier commissioning date Group for use of their libraries.
should be used in estimating old runner efficiency.
One comment on the peak efficiency equations for reac-
tion units is that the increase in efficiency due to size is References
based on the full Moody step-up formula with an exponent ASME. 1992. Performance test code for hydraulic turbines. Standard
of 5. It is common practice to use a combination of the PTC-18, American Society of Mechanical Engineers, New York.
Moody and Hutton formulae. The Hutton formula could not ASME. 1996. Hydropower mechanical design. American Society
be used because it requires knowledge of the model test of Mechanical Engineers, Hydro Power Technical Committee,
head, and this is not known. The full Moody step-up is used HCI Publications, Inc., Kansas City, Mo.
to allow for relatively lower peak efficiency in smaller run- Allis Chalmers. 1954. Hydraulic turbines and auxiliaries. Brochure
ners and the different manufacturing processes used with 02B7301, Allis Chalmers, York, Pa.
small runners, which are cast instead of fabricated and ma- Barrows, H.K. 1943. Water power engineering. 3rd ed. McGraw-
chined. Cast runners can have a much lower accuracy to the Hill Inc., New York, pp. 239–240.
model, thus resulting in lower efficiencies. Differences of Bhan, S., and Bhan, K. 1997. Rehabilitation of McCormick Units 1
just over 2% in peak efficiency have been recorded in identi- and 2. In International Water Power and Dam Construction,
cal cast runners which were subjected to Gibson efficiency Transactions, Uprating and Refurbishing Hydro Powerplants, 1–
tests, as was the case at Masson (Thicke and Gordon 1985). 3 Oct. 1997, Montréal, pp. 11–17.
Other equations have been developed to express the shape CEA. 1990. Hydroelectric turbine assessment for new units and run-
of the efficiency curve. However, the shape equations should ner retrofits. Canadian Electricity Association, Report 712 G 688.
only be used where there is no other available information. Creager, W.P., and Justin, J.D. 1949. Hydroelectric handbook. 2nd
For old runners, it is far preferable to undertake a relatively ed. John Wiley & Sons Inc., New York.
inexpensive index test to obtain the shape of the efficiency Davis, V.D. (Editor). 1952. Handbook of applied hydraulics. 2nd
curve, using the calculated peak efficiency, with an allow- ed. McGraw-Hill Inc., New York, pp. 564, 584.
ance for deterioration, to position the curve. The shape equa- Doering, J.C., and Gawne, K.D. 1998. Developing a traversing
tions appear to express the profile of the efficiency curve acoustic discharge measurement technique for the performance
with a reasonable degree of accuracy, mostly within ±1.5%, testing of low-head hydroelectric turbines. Canadian Journal of
Civil Engineering, 25: 777–778.
but occasionally deviating to around ±2% at lower flows. It
Doering, J.C., and Gawne, K.D. 2000. Acoustic discharge mea-
should be noted that the shape and peak efficiency of the ef-
surements for the performance testing of low-head hydroelectric
ficiency curve derived from equations for a Francis turbine turbines under disturbed flow conditions. Canadian Journal of
are based on far more data than those for other types of tur- Civil Engineering, 27: 160–165.
bines. Hence the efficiency curve for turbines other than Dumbleton, E.J. 1988. Upgrading hydro turbines: an operating au-
Francis should be regarded as preliminary, but accurate thority’s experience. International Water Power and Dam Con-
enough for feasibility studies. struction, 40(10): 11–14.
The peak efficiency and shape equations could also be Escher Wyss. 1981. Pelton turbines. Brochure e 21.16.30 ZC 81-
used as a benchmark, to determine whether a new unit has 20, Escher Wyss, Zurich, Switzerland.
the highest possible efficiency at the proposal stage, or to de- Gibson, N.R. 1958. Experience in the use of the Gibson method of
velop an efficiency curve for a unit at the feasibility study water measurement for efficiency tests of hydraulic turbines.
stage, where there is no information available from manufac- Transactions of the ASME, Journal of Basic Engineering, Paper
turers. By combining these efficiency equations with unit No. 58–A-78.
sizing and setting formulae developed for the CEA (1990), Gilbert Gilkes and Gordon. 1973. The Turgo impulse turbine. Bro-
including the manufacturing technology revisions (Gordon chure 5M/2/73 WY, Gilbert Gilkes and Gordon, Water turbine
1999), it is now possible to develop computer programs manufacturers, Kendall, England.
which can produce an accurate estimate of runner size and Gordon, J.L. 1970. The Harca hydro development. Transactions,
efficiency. Engineering and Operating Division, Canadian Electrical Asso-
ciation, 8-2.
Gordon, J.L. 1999. A statistical method for determination of hy-
Conclusions draulic turbine size and setting. In Canadian Electricity Associa-
This analysis has shown that it is possible to develop em- tion, Spring Meeting, Vancouver, 1999.
Grein, H., Klicov, D., and Wieser, W. 1988. Efficiency scale effects
pirical equations from statistical data to determine the peak
in Pelton turbines. International Water Power and Dam Con-
efficiency and the shape of the efficiency curve for both new
struction, 40(5): 32–36.
and old hydraulic turbines. The equations should prove to be
Grunder, R., Brandt, R., and Panzenbock, R. 1997. Increase of
most useful in feasibility studies and as an aid in estimating maximum discharge in Kaplan turbines by CFD-optimized new
old turbine performance where there are no other data avail- runners. In International Water Power and Dam Construction,
able. Transactions, Uprating and Refurbishing Hydro Powerplants, 1–
3 Oct. 1997, Montréal, pp. 221–224.
Acknowledgements Gujan, H., and Hauser, H. 1994. New Pelton runners for the Kublis
hydroelectric powerplant. International Water Power and Dam
The author wishes to thank Great Lakes Power Ltd. for Construction, 46(2): 26–28.
providing access to their files, Newfoundland and Labrador Hackert, H. 1976. Victor Kaplan — his life and work. International
Hydro for providing data on Cat Arm efficiency, and the Ca- Water Power and Dam Construction, 28(11): 39–41.

© 2001 NRC Canada

Gordon 253

Holzer, W., and Schedler, E. 1994. Upgrading of Latschau power Thicke, R.H., and Bakar, L.J. 1972. Plant layout and turbine design
station: Kaplan turbines. International Water Power and Dam of the Bighorn Power Project. Canadian Electrical Association,
Construction, 46(2): 28–29. Transactions, Engineering and Operating Division, 10-2.
Jungbauer, G., Hegenbarth, A., and Grunder, R. 1992. Runner re- Thicke, R.H., and Gordon, J.L. 1985. Hydro turbine rehabilitation.
placement at Losenstein hydroplant. Brochure 29.06.30 ZC-10, Canadian Electrical Association, Transactions, Engineering and
Sulzer Escher Wyss, Zurich, Switzerland. Operating Division, 23-2.
Kvaerner Energy. 1997. Turbine modernization and uprating. Bro- USBR. 1976. Selecting hydraulic reaction turbines. United States
chure KE 016/5/97, Kvaerner Energy, Oslo, Norway. Department of the Interior, Bureau of Reclamation, Engineering
Kepler, J.L., and Mallam, J. 1997. Runner upgrade at the Bay Monograph 20.
d’Espoir generating station. In International Water Power and Vitvar, M. 1989. Kaplan turbine uprating at the Orlik powerplant.
Dam construction, Transactions, Uprating and Refurbishing Hy- International Water Power and Dam Construction, 41(10): 24–26.
dro Powerplants, 1–3 Oct. 1997, Montréal, pp. 19–23. Voith Hydro Inc. 1992. Voith Hydro continues as major turbine
Lambert, C., and Jarriand, P. 1995. Grand Rapids turbine upgrad- supplier. Forum, Vol. 3, pp. 4–5. Voith Hydro Inc., York, Pa.
ing and repair of unit. Transactions, Engineering and Operating Vuillioud, G., and Walther, W. 1988. Rehabilitation and power in-
Division, Canadian Electrical Association, 33-2. crease of the hydroelectric facilities at Verbois. Vevey Technical
Lang, J.H., and Thomas, D.R. 1996. Kootenay Canal turbine up- Bulletin 88, Vevey Engineering Works, Vevey, Switzerland,
grade, homology to field tests. Transactions, Engineering and pp. 40–48.
Operating Division, Canadian Electrical Association, 34(Part 2). Wachter, G.F. 1998. Creating new life for hydraulic turbines by up-
Lowy, R. 1949. Speed regulation computations for hydraulic turbines. grading and rehabilitation. Canadian Electric Association, Elec-
In Proceedings of the American Society of Mechanical Engineers tricity ‘98, Toronto.
Annual Meeting, New York, Paper 48-A-37, Figs. 7 and 8.
Markovich, M., Yap, R., Jermann, P., and Pomeroy, K. 1991. R.H.
Saunders G.S. — Propellor turbine runner replacement project
outline and model development. In Transactions of Waterpower List of symbols
‘91, Vol. 3, pp. 1946–1954. d turbine throat diameter in metres
Mahe, B., and Miron, J-G. 1991. LG2A turbine-generator units, djet jet diameter in metres
evolution in design, manufacture and installation. Transactions,
h net head on turbine in metres
Engineering and Operating Division, Canadian Electricity Asso-
hrated turbine rated head in metres
ciation, 29-2, Fig. 2.
j number of jets on an impulse turbine
McLaughlin, B. 1994. Upgrading an old facility. Transactions, En-
gineering and Operating Division, Canadian Electrical Associa- k flow exponent
tion, 32-2. nq metric specific speed based on flow; for impulse
Moody, L.F. 1952. Hydraulic machinery. In Handbook of applied units, one jet flow
hydraulics. 2nd ed. Edited by C.V. Davis. McGraw-Hill Inc., Q turbine flow in m3/s
New York, pp. 599, 603. Qjet impulse turbine flow per jet at rated head in m3/s
Nestorov, V., and Pavleski, V. 1997. Revitalization and moderniza- Qpeak turbine flow at peak efficiency and rated head
tion of the hydro powerstation “Spilje.” In International Water in m3/s
Power and Dam Construction, Transactions, Uprating and Re- Qpj jet flow at peak efficiency and rated head in m3/s
furbishing Hydro Powerplants, 1–3 Oct. 1997, Montréal, Qrated turbine flow at rated head and rated load in m3/s
pp. 47–55. Qsnl turbine flow at speed-no-load and rated head
Nichtawitz, A. 1997. Influence of surface roughness to hydraulic in m3/s
performance. In International Water Power and Dam Construc- rpm turbine synchronous speed in revolutions per min-
tion, Transactions, Uprating and Refurbishing Hydro Power- ute
plants, 1–3 Oct. 1997, Montréal, pp. 139–146. y year unit commissioned, with latest date being 1998
Pfafflin, G.E. 1987. Optimizing the yield from existing hydro- De change in turbine efficiency (decimal)
electric generating facilities. Pulp and Paper Canada, 88(11): De head change in turbine efficiency due to change in head
402–409. from rated head (decimal)
Rogers, F.H., and Moody, L.F. 1925. Inter-relation of operation and De peak change in turbine efficiency from peak efficiency
design of hydraulic turbines. In Proceedings of the 3rd Ameri-
(decimal)
can Annual Hydro-Electric Conference, Engineers Club of Phil-
De size change in turbine efficiency (Moody formula) due
adelphia, 10 Mar. 1925, pp. 1–18.
to size (decimal)
Sabourin, M., Eremeef, R., and De Henau, V. 1995. Extensive use
of computational fluid dynamics in the upgrading of hydraulic De specific speed change in turbine efficiency due to specific speed
turbines. Canadian Electrical Association, Transactions, Engi- (decimal)
neering and Operating Division, 33-2. De nq change in turbine efficiency due to metric specific
Sebestyen, A., and Keck, H. 1995. Uprating of ultra low head speed (decimal)
Francis units based on numerical flow analysis. In International De year change in turbine efficiency due to year unit com-
Water Power and Dam Construction, Transactions, Uprating and missioned (decimal)
Refurbishing Hydro Power Plants, 9–11 Oct. 1995, Nice. DQ change in flow from peak efficiency flow (decimal)
Sulzer Hydro. 1996. Runner replacements in hydropower plants: a e peak turbine peak efficiency (decimal)
new energy source. Brochure e/21.68.30-ZA 96-075, Sulzer Hy- e pj impulse turbine jet peak efficiency (decimal)
dro, Kriens, Switzerland. e q turbine efficiency at flow Q (decimal)

© 2001 NRC Canada