Beruflich Dokumente
Kultur Dokumente
P J Purcell
Dooge Centre for Water Resources Research,
University College Dublin, Eire
ABSTRACT
1 INTRODUCTION
The objective of the present paper is to describe a transient event that occurred in 2009 at
a medium-sized hydro-electric scheme and the subsequent modelling of this event.
Although, there are a number of generic software packages available for modelling fluid
transients, a specific computational model, that more faithfully represented the boundary
conditions of the system, was developed. In addition, one of the key issues in this case
study was the development of a model that reasonably represented the system with
limited measured data. The numerical model of the fluid transients has been developed
using an EXCEL spreadsheet, enabling ‘what-if’ scenarios to be explored arising from
various operating regimes at the hydroelectric scheme. At present, there is no measured
waterhammer data available to verify the model described below, thus the issue of
validating the model is explored below.
2 LITERATURE REVIEW
Surge tank
Area A 2
+Z
hf
Turbines
Much has been published on the performance of surge tanks in full-scale installations. As
far back as 1933, Gibson et al. compared the results of a simple surge tank on a full-scale
hydro-electric sysem at Tallulah Falls, Georgia with the numerical predictions (2).
Goodhue, H.W. in 1943 compared the performance of a differential surge tank at
Apalachia, Tennessee with the computational results (3). Bratfisch et al. in 1955 showed
good agreement between calculated pressure transients at the Owens Gorge Power Plant
using the graphical method and the test data measured with calibrated Bourdon gauges
(4). In 1963, Pearsal undertook a series of experiments to evaluate the performance of
various types of laboratory-scale surge tank and also compared a number of
computational methods with measured data (5). Noguueira in 1989 compared numerical
and physical models of surge tank performance and concluded that physical model
studies of surge tanks could be conducted using Froude similarity (6). Tănasel et al. in
2010 analysed the free surface level variation in a surge tank and in the secondary shafts
of a complex hydro electrical scheme with two headraces and four secondary intakes (7).
Vereide commenced a 4-year project in 2012 to undertake in-situ measurements of
existing surge tanks in Norway to develop, calibrate and validate physical and numerical
models of these surge tanks (8). In addition, the optimization of surge tank throttling is
being examined as part of this research project.
19.96 m
189.6
T.W.L. 186.3
167.3
398 m
Inlet valves
Turbines
4 MODEL DEVELOPMENT
The starting point in the development of a model of the hydroelectric scheme was the
formulation of a spreadsheet model of the classical reservoir-pipe-turbine system,
protected by a surge tank, as shown schematically in Fig. 3.
Surge tank
Area A2
+Z
hf
Turbines
Notation:
L = pipeline length
V = fluid velocity
A1 = cross-sectional area of pipe
A2 = plan area of surge tank
g = acceleration due to gravity
hf = head loss due to friction (fLv2/2gd)
z = water level in surge tank
t = time following initiation of transient event
Qturbine = flow through turbines
In the hydroelectric system shown in Fig. 3, under steady operating conditions, the water
level in the surge tank will be at an elevation hf below the reservoir level due to friction
and fitting losses. Consider the sudden closure of the control valves in this system. As the
valves close, the flow along the pipeline is directed up into the surge tank, resulting in a
rise in elevation of the water surface in the surge tank. As the water level in surge tank
rises above the reservoir level, the column of water in the pipe decelerates, eventually
reaching zero velocity. The out-of-balance head between the surge tank and the reservoir,
at the time of zero water velocity in the pipeline, results in the fluid column reversing in
The two equations describing this transient motion through the system are the
conservation of fluid momentum and conservation of fluid mass (continuity). The
momentum equation describes how the deceleration (dv/dt) of the fluid in the pipeline
that occurs as a result of the friction/fitting losses (hf) and the rise/fall in the elevation of
the free water surface in the surge tank (z). The continuity equation simply states that,
under transient flow conditions, flow through turbines is simply the sum of the flow
along the pipeline plus the flow from the surge tank.
Momentum equation
ௗ
ௗ௧
ݖേ ݄ ൌ Ͳ (1)
Continuity equation
ௗ
ܣଵ ܸ ൌ ܣଶ ௗ௧ ܳ௧௨ (2)
Exact solutions to these equations are only possible for simple boundary conditions. For
more realistic situations, approximate solutions can be developed by numerical methods
such as finite difference formulations of equations (1) and (2). Solution of equation (1)
by finite difference yields the water velocity in the intake tunnel at time t + 't:
ௗ
V t + 't = V t +
ௗ௧
߂ 1(a)
Solution of equation (2) by finite difference yields the water level in the surge tank at
time t + 't:
ௗ
Z t + 't = Z t + ௗ௧
߂ 2(a)
The errors associated by the use of first order methods such as the Euler scheme given by
equations 1(a) and 2(a) can be minimised by using an appropriate computational time
step or alternatively employing higher-order methods such as the Runge-Kutta ( 9).
5 MODEL DEVELOPMENT
The transient water level in the surge tank computed by model developed for this case
study is shown in Fig. 4. A comparison of the textbook values and the computed values
from the developed excel model is shown in Table 1.
Fig. 4 Computed water level in surge tank for textbook case [10]
Fig. 5 (b) Water level variation in surge tank at hydroelectric scheme for a sudden
load rejection by turbines operating at 80 m3/s, assuming no friction, time step
0.125s
189.6
W.L. 187.05
Surge tank
Reservoir Gate Air vent
167.3 39 m3 /s
Inlet valves
Turbines
137.05
Fig. 6 Water levels and flows prior to transient event at hydroelectric scheme
The excel spreadsheet model described above that was developed for the ‘standard’
textbook reservoir-pipe-surge tank, for a transient event arising from load rejection by the
turbines, was modified to model more accurately the actual transient event that occurred
at the hydroelectric scheme, as follows:
(i) Inclusion of air vent, adjacent to reservoir, in model;
(ii) Initiation of transient event by closure of the intake gate.
Since, in this case, a transient event resulted from the sudden deceleration of the flow
through the gate, it is important to consider the parameters influencing the discharge
through the intake gate. The discharge through the gate at any given time (t seconds)
following initiation of gate closure is a function of:
(i) the head of water overlying the gate;
(ii) the coefficient of discharge for the gate;
(iii) the flow area beneath the gate at time t.
Following sudden closure of the intake gate, the water column in the air vent is drawn
down and flows into the tunnel to try to compensate for the loss of flow due to the gate
closing. As the out-of-balance head between the water level in the surge tank and the air
vent increases, the water column in the tunnel decelerates, reaches zero velocity and
eventually reverses in direction. During the flow reversal, the outflow from the surge
tank bifurcates through the turbines (if still open) and into the tunnel. The model assumes
the worst case scenario, that is, all outflow from the surge tank discharges into tunnel
causing the largest upsurge through the air vent adjacent to the intake gate.
7 COMPUTATIONAL OUTPUT
On the day of the transient event, the intake gate was accidentally tripped, causing an
inflow of water from the air vent shaft and a surge in the pressure tunnel. One turbine
was operating at the time, at an approximate flow of 39 m3/s. It is estimated that the
intake gate closed within 15 – 22 seconds and that water vented from the air intake in a
time interval 22s - 30s after the intake gate was tripped. Anecdotal evidence is that water
discharged from the air vent to about window height in the adjacent intake house.
The computational output illustrating the transient water levels in the air vent following
closure of the intake gate is shown in Fig. 7. Note that all levels shown are relative to the
steady water level prior to the transient event (187.05m). Following closure of the gate,
the water level in the air vent falls to a level of about 8m below the initial level. The
periodic time of the mass oscillation in the tunnel is about 14s and the computational
model suggests that a reflection of the wave from the surge tank occurs before closure of
gate at a time of 20s. Following closure of the gate, there is a sharp increase in the water
level in the air vent, reaching a peak height of about 7m above the initial steady water
level in the air vent or about 4.5m above ground level at the air vent at a time of about 28
seconds after the start of gate closure. This computational output correlates reasonably
well with the findings of the report into incident.
The model was re-run for slower gate closure times with one turbine running at 39 m3/s.
Figures 8, 9 and 10 show the computed transient water levels in the air vent shaft for gate
closure times of 40s, 60s and 120s respectively. Examination of the figures shows that
slowing the gate closure reduces considerably the magnitude of the peak water level and
delays its time its time of occurrence. For the water to discharge above ground level from
the air vent shaft, the transient peak water level would have to reach a height of about
2.5m above its initial steady level and clearly, in the case of figures 8, 9 and 10, this does
not occur.
Fig. 7 Water level in air vent relative to steady level, closure of intake gate in 20s –
steady flow 39 m3/s
Fig. 8 Water level in air vent relative to steady level, closure of intake gate in 40s –
steady flow 39 m3/s
Fig. 9 Water level in air vent relative to steady level, closure of intake gate in 60s –
steady flow 39 m3/s
Fig. 10 Water level in air vent relative to steady level, closure of intake gate in 20s –
steady flow 39 m3/s
Fig. 12 Water level in air vent relative to steady level, closure of intake gate in 40s –
steady flow 80 m3/s
Fig. 13 Water level in air vent relative to steady level, closure of intake gate in 60s –
steady flow 80 m3/s
Fig. 14 Water level in air vent relative to steady level, closure of intake gate in 120s
– steady flow 80 m3/s
8 CONCLUSIONS
An incident resulting from the accidental closure of the intake gate to the penstock of a
hydroelectric scheme has been presented above. A numerical model of this incident and
The computational model has been validated with other published data and by
comparison with analytical solutions for simple boundary conditions. While no measured
transient data was available in this case, anecdotal evidence from the transient incident
has been used to corroborate the computational output. This case study illustrates the
need to take pro-active measures such as the prevention of the accidental sudden
reduction of the flow to the turbines (e.g. preventing sudden intake closure) and the
installation of physical mechanisms to control the rate of gate closure in the event of
accidental tripping (e.g. pneumatic actuator on intake gate to control its rate of closure).
REFERENCES