Penstock failure detection system at the "Valsan" hydro power plant

Abstract

"Valsan" is a small Hydro Power Plant, 5 MW, situated at about 160 km
north of Bucharest, Romania, on the small "Valsan" river in a remote
mountainous area. It is equipped with a single Francis turbine. The
penstock is located in the access shaft of the HPP. "Hidroelectrica",
the Romanian company that operates the HPP, was trying to implement a
remote penstock failure detection system. Starting from a classic
hydraulic problem, the authors of the paper derived a method for failure
detection and localization on the pipe. The method assumes the existence
of 2 flow meters and 2 pressure transducers at the inlet and outlet of
the pressurized pipe. Calculations have to be based on experimental
values measured in a permanent regime for different values of the flow
rate. The method was at first tested on a pipe, in the Hydraulic
Laboratory of the Technical University of Civil Engineering Bucharest.
Pipe failure was modelled by opening of a valve on a tee branch of the
analyzed pipe. Experimental results were found to be in good agreement
with theoretical ones. The penstock of the "Valsan" HPP, was modelled in
EPANET, in order to: i) test the method at a larger scale; ii) get the
right flow and pressure transducers that are needed to implement it. At
the request of "Hidroelectrica" a routine that computes the efficiency
of the turbine was added to the monitoring software. After the system
was implemented, another series of measurements were performed at the
site in order to validate it. Failure was modelled by opening an
existing valve on a branch of the penstock. Detection of the failure was
correct and almost instantaneous, while failure location was accurate
within 5% of the total penstock length.

Full text

Penstock failure detection system at the "Valsan" hydro power plant
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
2012 IOP Conf. Ser.: Earth Environ. Sci. 15 052005
(http://iopscience.iop.org/1755-1315/15/5/052005)
Download details:
IP Address: 193.231.3.27
The article was downloaded on 11/01/2013 at 12:55
Please note that terms and conditions apply.
View the table of contents for this issue, or go to the journal homepage for more
Home Search Collections Journals About Contact us My IOPscience

Penstock failure detection system at the “Valsan” hydro
power plant
A M Georgescu1, C I Coşoiu1, N Alboiu1, D Hlevca1, R Tataroiu2 and O Popescu3
1 Hydraulics and Environmental Protection Department, Technical University of Civil
Engineering Bucharest, B-dul Lacul Tei, Nr. 124, Sector 2, 020396, Bucharest,
Romania
2 Computer Science Department, Politehnica University of Bucharest, Splaiul
Independentei, Nr. 313, Sector 6, Bucharest, Romania
3 Sangari Engineering Services – Romania, Str. Ripiceni, Nr. 2, Bl. 12, Sc. C, Ap. 48,
Sector 2, 023623, Bucharest Romania
E-mail: andrei_georgescu2003@yahoo.com
Abstract. “Valsan” is a small Hydro Power Plant, 5 MW, situated at about 160 km north of
Bucharest, Romania, on the small “Valsan” river in a remote mountainous area. It is equipped
with a single Francis turbine. The penstock is located in the access shaft of the HPP.
“Hidroelectrica”, the Romanian company that operates the HPP, was trying to implement a
remote penstock failure detection system. Starting from a classic hydraulic problem, the
authors of the paper derived a method for failure detection and localization on the pipe. The
method assumes the existence of 2 flow meters and 2 pressure transducers at the inlet and
outlet of the pressurized pipe. Calculations have to be based on experimental values measured
in a permanent regime for different values of the flow rate. The method was at first tested on a
pipe, in the Hydraulic Laboratory of the Technical University of Civil Engineering Bucharest.
Pipe failure was modelled by opening of a valve on a tee branch of the analyzed pipe.
Experimental results were found to be in good agreement with theoretical ones. The penstock
of the “Valsan” HPP, was modelled in EPANET, in order to: i) test the method at a larger scale;
ii) get the right flow and pressure transducers that are needed to implement it. At the request of
“Hidroelectrica” a routine that computes the efficiency of the turbine was added to the
monitoring software. After the system was implemented, another series of measurements were
performed at the site in order to validate it. Failure was modelled by opening an existing valve
on a branch of the penstock. Detection of the failure was correct and almost instantaneous,
while failure location was accurate within 5% of the total penstock length.
1. Introduction
Valsan is a small HPP that produces 5 MW using the energy from the Valsan Lake. The total volume
of the Valsan lake is about 175 000 cubic meters from which the usable volume is about 120 000 cubic
meters for a height of 5.50 m from the maximum level (+954.50 m above sea level) to the minimum
one (949.00 m above sea level).
The headrace tunnel (see figure 1) is built of reinforced concrete and has the external diameter of
1.8 m, the internal diameter of 1.4 m and the total length of 160 m. A surge tank is placed at the lower
end of the headrace tunnel and it has a height of 17.5 m.
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 052005 doi:10.1088/1755-1315/15/5/052005
Published under licence by IOP Publishing Ltd
1

A butterfly valve is placed at the lower end of the headrace tunnel after the surge tank. The valve
has a diameter of 1.4 m and is isolating the penstock from the headrace and the lake. If the operational
butterfly valve is failing to close then this valve is closing the circuit. This valve is also used to close
the hydraulic circuit when maintenance is performed on the operational butterfly valve. In the case of
protecting the HPP from floods or in the case of penstock failure this valve is also used to close the
circuit.
The penstock pipe is connecting the headrace pipe with the turbine and it has the internal diameter
of 1.2 m and a length of 100 m. At a level of 852.799 m above sea level the penstock bifurcates: one
pipe connects to the operative valve and the other is used to by-pass the turbine going directly to the
tailrace. Both pipes have a 1.2 m diameter.
The Valsan HPP has a horizontal double rotor Francis turbine of 5.3MW and the rotation speed of
500 rot/min, which is coupled to an electric generator of 6.2 MVA. The operation butterfly valve is
situated between the turbine and the bifurcation of the penstock. This is a valve that can be
automatically operated to permit the water inlet towards the turbine.
One of the emergency problems that can appear at the HPP is the breakdown of the penstock.
Figure 1. The Valsan hydro power plant (HPP)
2. Problem statement and theoretical approach
As the penstock is located in the access shaft of the HPP the request from “Hidoelectrica” was for a
penstock failure detection system that would be able to detect breakdown, signal it as soon as possible
and compute its location along the penstock as accurately as possible. Except for the time delay
between the failure and its detection, this sounded almost exactly like a classical hydraulics problem
[1]: A flow meter and a pressure transducer are positioned at each end (the upstream one and the
downstream one) of a straight horizontal pipe of known length, diameter and Darcy friction factor.
The fluid is considered to be incompressible, of known density and kinematic viscosity. In case of
breakdown, given the values of the flow rate and pressure at each of the ends of the pipe compute the
location of breakdown.
Assuming that the flow at both ends of the pipe is a fully turbulent rough flow, the Darcy friction
factor is constant with respect to the Reynolds number (i.e. with respect to the flow rate). As the pipe
is straight the minor head losses are sensibly smaller than the friction head losses and can be neglected
[2]. Moreover the kinetic terms from the energy equation can also be neglected (they have generally
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 052005 doi:10.1088/1755-1315/15/5/052005
2

the same magnitude order as the minor head losses). In this academic problem there are no level
differences between the ends of the pipe as it is horizontal.
Denoting the position of the breakdown measured from the upstream end of the pipe as
x
and
considering all the above mentioned assumptions the energy equation reduces to:
g
v
DxL
g
v
D
x
g
pp
down
updownup
2
)(
2
2
2
−
+=
−
λλ
ρ
(1)
where the subscripts “up” and “down” stand for upstream and down stream respectively,
p
is the
pressure,
ρ
the fluid density,
g
the acceleration of free fall,
λ
is Darcy’s friction factor,
D
the
diameter of the pipe,
L
the length of the pipe and
v
the velocity (which can be computed for an
incompressible fluid knowing the volumetric flow rate “
Q
” and the diameter of the pipe as . 2
4
D
Q
v
π
=)
This is a first degree equation that can be easily solved for
x
.
In a real case, like the one of the penstock, some of the assumptions made for the academic
problem are no longer valid. There are level differences between the two ends of the pipe. Although
the flow regime can be assumed fully turbulent rough flow in most of the cases, the Darcy friction
factor is not known. There might be minor head losses along the pipe. There are also other questions
that should be answered in a real case. What would the accuracy of such a method be? This method
could give good results in steady flow but it is unlikely it would do the same in unsteady flow
conditions [3].
Of course, in order to assess those problems, the theoretical model should firstly be changed in
order to handle the different assumptions that appear in a real case. First, instead of using the pressure
differences between the upstream and downstream ends of the penstock, in the real case, the difference
of piezometric heads should be used (
p
Hp z g
ρ
= +
.). Second, the head losses must be computed
correctly. When writing the energy equation for the penstock (when there is no breakdown
up down
QQ Q= =
) in terms of flow rate we obtain:
2
42
2
16 Q
Dg
D
L
HpHp
downup
π
ζ
λ





+=− ∑
(2)
Denoting by
M
(resistance modulus of the penstock) the term





+∑
ζ
λ
π
D
L
Dg
42
8
and by
Hp∆
the difference between the piezometric heads of the ends of the pipe, equation (2) becomes:
2
MQHp =∆
(3)
A set of several measurements performed on the pipe, in steady flow conditions, without
breakdown, would permit the correct calculation of the resistance modulus of the pipe with a least
squares algorithm as:
( )
∑
∑
∆
=
4
2
i
ii
Q
QHp
M
(4)
Now, with a known value of the resistance modulus of the pipe, considering the minor head losses
small enough to be neglected, equation (1) becomes:
( )
22 downupdownup
QxL
L
M
xQ
L
M
HpHp −+=−
(5)
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 052005 doi:10.1088/1755-1315/15/5/052005
3

As stated above this equation is only valid as long as there are no minor head losses on the
penstock or as long as the minor head losses are small enough to be neglected. In fact, as in equation
(5) we are dealing with
L
M
, the two terms that should be compared are: on one hand,
D
λ
and on the
other hand
L
∑
ζ
. For the existing penstock, at the Valsan H.P.P. the orders of magnitude are: for
λ
-
2
10−
, for
D
-
0
10
, for
L
-
2
10
and for
∑
ζ
-
1
10−
. The result of the comparison shows that the term
L
∑
ζ
is one order of magnitude smaller than
D
λ
which implies that equation (5) could be used
directly. Of course, the results will be approximate values for the location of the breakdown but a
more accurate equation would be difficult to implement in a monitoring and detection computer
program. The results of implementing equation (5), for an existing pipe, with minor head losses, were
tested in a laboratory set-up.
3. Laboratory set-up and tests
As stated above the aim of the laboratory tests was to check the accuracy of equation (5) for
breakdown location in the case of a pipe that has local head losses.
For the experiments, a high pressure network existing in the Hydraulics and Environmental
Protection Laboratory of the Technical University of Civil Engineering Bucharest was used.
The network is presented in fig. 2. The equipment includes: a tank, a high pressure pump, a 100
mm diameter steel pipe (with two 90o bends and a tee branch that was used to simulate leakage), a
differential pressure manometer, two ultrasonic flow meters and an electromagnetic flow meter. The
electromagnetic flow meter was used to calibrate the ultrasonic flow meters before the experiments.
Figure 2. Laboratory experimental set-up; DPG – Differential pressure manometer; EFM –
Electromagnetic flow meter; HPP – High pressure pump; UFM – Ultrasonic flow meter; Vc; Vd; Vu -
Valves
The experimental tests were performed in two stages. The first stage was to establish the value of
the resistance modulus of the pipe. Therefore the valve Vu was completely opened, valve Vc was
completely closed and valve Vd was used for adjustment of the flow rate through the system. The
parameters of interests were the flow rate and the head loss
Hp∆
between the upstream and the
downstream measurement cross sections. Based on equation (4), the value of the resistance modulus of
the pipe was determined. A graph showing the measured and computed values for equation (3) was
plotted in figure 3. The value of the computed resistance modulus for the pipe was M = 19370.89679
s2/m5.
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 052005 doi:10.1088/1755-1315/15/5/052005
4

Figure 3.
Computed and experimentally obtained values for the
resistance modulus of the pipe
The second stage of the laboratory tests aimed at locating the breakdown of the pipe using equation
(5)]. In order to simulate the pipe breakdown, the Vc valve was opened. Three opening degrees for
valve Vc were used in the experiments, namely ¼, ½ and ¾. The differences in piezometric heads for
different flow rates, obtained by adjusting the Vd valve, were recorded and the location of the
breakdown vas calculated according to equation (5).
The distance between the upstream measuring cross section and the tee branch is 12.76 m. The
computed locations of the breakdown laid between 10 and 20 m for all cases. As expected, the results
provided by equation (5) are approximate locations of the breakdown. Nevertheless errors in the
estimated locations with respect to the total length of the pipe are in the range of
±
5%. This was
considered to be an acceptable range for the problem at hand.
4. Numerical tests and choices for measuring devices
The penstock of the “Valsan” HPP, was modeled in EPANET, a free software produced by the
Environmental Protection Agency of the U.S., using a series of experimental values measured at the
site in order to: i) test the method at a larger scale; ii) get the right flow and pressure transducers that
are needed to implement it.
The geometric values used for the EPANET model were the exact geometric dimensions existing in
the “Valsan” HPP, the Darcy-Weisbach formula, implemented in EPANET was used to compute the
friction head losses for an absolute roughness coefficient of the pipe of 2 mm. Local head loss
coefficients were added for the curves and the tee branch of the penstock. The exit nodes of the pipes
(turbine and by-pass) were considered as emitters (i.e. nodes that discharge water at atmospheric
pressure). The only parameters that were modified during the simulations were the local head losses
coefficients of the two valves (turbine and by-pass, see figure 4). All tests were performed for steady
state situations [4]. The same procedure as in the laboratory test was used: we performed preliminary
tests with the by-pass valve closed in order to compute the resistance modulus of the penstock; after
the value of the resistance modulus was assessed another series of tests were performed with different
openings of the by-pass valve.
The computed value of the resistance modulus of the penstock was 0.15 s2/m5. For the breakdown
location tests, the accuracy of the estimation laid in the same range as the tests performed in the
laboratory (as the local head losses included in the numerical model were neglected in the theoretical
model).
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 052005 doi:10.1088/1755-1315/15/5/052005
5

Figure 4. EPANET model of the Valsan HPP
The maximal flow rate recorded during numerical tests was of approximately 7 m3/s, on the
penstock, before the tee branch. The maximal pressure on the downstream end of the penstock was of
about 11 bars, while on the upstream end it was only of about 0.9 bars.
These values were used to choose the pressure transducers and the flow meters for the “Valsan”
system.
The flow meters are 2 identical GE Panametrics ultrasonic flow meters, model AquaTrans AT868,
with 2 pairs of GE Panametrics line for ultrasonic flow transducers. These ultrasonic flow transducers
are mounted at the exterior of the pipe and measure the flow rate of sonically-conductive liquids
through pipes having diameters between 5 cm and 7.6 m. The transducers are mounted on opposite
sides of the pipe so the measurements are performed with only one pass of the sound wave.trough the
pipe. The supports of the transducers are welded to the pipe. The measurements are typically
independent of the pipe material.
The pressure transducers are 2 Endress&Hauser programmable pressure transducers, model
Cerebar S PMC71. This type of transducers is capacitive with a Ceraphire membrane. The accuracy is
of 0,075% from the programmed scale and they are resistant to high pressure shocks (x10 full scale).
The output is a unified 4-20 mA signal. The 2 pressure transducers have the following measuring
ranges: for the upstream end of the penstock, 0-2 bars (used at full scale); for the downstream end of
the penstock, 0-40 bars (programmed on site for a 0-12 bars scale) [5].
Data acquisition and control is based on National Instruments hardware and software: CompactRIO,
(NI cRIO-9012) real-time controller with 64 MB DRAM, 128 MB storage, 400 MHz processor,
embedded controller running LabVIEW Real-Time, for deterministic control, data logging, and
analysis; 16-Bit Analog Current Input Module (NI 9203), 8-analog input channels ±20 mA, 200 kS/s
sampling rate.
5. On site software calibration and tests
The software for failure detection was created in LabVIEW. It acquires data from the 4 sensors at a
sampling rate of 0.5 Hz. In addition to the 4 sensors, the state of the operational valve (i.e. opened;
closed or none, the latter corresponding to the valve being in the process of opening or closing) is also
acquired together with the electric power output of the turbine. At the request of the beneficiary, a
simple routine was added to the software. It computes the overall efficiency of the turbine by
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 052005 doi:10.1088/1755-1315/15/5/052005
6

considering as useful power the electric power output of the turbine and as consumed power the power
of the fluid (
gQH
ρ
- where the flow rate represents the mean value of the flow rates measured by the
2 ultrasound flow meters and the head is computed as the difference between water level at the
reservoir and the level of the end of the pipe towards the tail race). The state of the operating valve is
used by the software to distinguish between steady and unsteady flow in the penstock. During the tests
we performed, we noticed that when the flow was unsteady (i.e. when the operating valve was in the
process of opening or closing) a difference of flow rates between the upstream and downstream cross
sections was recorded and this difference could, under certain conditions, trigger the breakdown alert.
After the state of the operating valve was acquired we disabled the alert while the valve is moving and
for a 2 seconds lag after the valve has stopped at the desired position.
When the software was completed, a series of tests were performed in steady flow to determine the
value of the resistance modulus of the penstock. The measured values and the computed curve are
presented in figure 5. The computed value of the resistance modulus of the penstock (of 0.141 s2/m5)
was added to the software in the sequel and another series of tests were performed in unsteady flow in
order to insure there are no false alerts of breakdown. Finally a third series of tests that implied the
opening of the by-pass valve were performed in order to quantify the accuracy of the breakdown
location detection and response time.
Figure 5.
Computed and experimentally obtained values of the
resistance modulus for the Valsan Penstock
Breakdown location was computed by the software with a
%5±
error with respect to the total
length of the penstock. The full response time for (i.e. breakdown detection and location) was of about
one minute. We have to mention that the breakdown is detected less than 10 seconds after the by-pass
valve was opened more that 10% but correct location detection appears only after about one minute. In
the first seconds after the breakdown was detected, its location, computed by the software, is
inaccurate (varying between
%25±
of the actual location with respect to the full penstock length).
This is due to two major reasons. First of all, the by-pass valve is a manual valve that requires about
one minute to be fully opened or closed. During this process high unstable flow are recorded in the
penstock. Second, high unsteady flows are expected for some seconds at the beginning of the
breakdown process. In other words, the time lag recorded between detection and correct location of
the breakdown in a real case will still exist but is expected to be smaller than the one recorded during
the experiments (as the unsteady flows due to the opening time of the by-pass valve will no longer
exist in a real case). All experiments performed for openings of the by-pass valve smaller than 10%
were not accurate as long as the location of the breakdown is concerned. Correct detection of the
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 052005 doi:10.1088/1755-1315/15/5/052005
7

breakdown starts from about 5% opening of the by-pass valve. As a general rule observed during these
tests, the greater the opening of the by-pass valve, a more accurate detection of the location occurs.
After all the tests were performed another condition was added to the software. Any breakdown
detection located within
%5±
of the actual position of the by-pass pipe on the penstock will not be
reported. This last condition was obviously needed so that the system will not launch an alert when the
by-pass valve is opened.
6. Conclusions
Starting from a simple hydraulics problem a system was developed for detection and location of a
breakdown in the penstock of a small HPP in Romania. Although not extremely accurate, especially
for small values of the flow rate trough the breakdown, and working accurately only in steady flow
conditions the system was implemented at the HPP of “Valsan”. It is working since April 2011 and up
to this date has not launched any false breakdown alerts. Flow rate measurements in the penstock also
permitted integration of a routine computing the actual overall efficiency of the HPP. Extensive tests
(that are generally very hard to perform) are needed in order to evaluate the application of this system
to a more important HPP.
References
[1] Isbasoiu E C and Georgescu S C 1995 Mecanica Fluidelor (Bucharest, Ed. Tehnica) p 351
[2] Georgescu A M and Georgescu S C 2007 Hidraulica Retelelor de Conducte si Masini
Hidraulice (Bucharest, Ed. Printech) pp 42-44
[3] Adamkowski A 2012 Discharge Measurements Techniques in Hydropower Systems with
Emphasis on the Pressure-Time Method in Hydropower – Practice and Applications (Croatia:
InTech) pp 94-98
[4] ISO 5168 2005 Measurement of Fluid Flow – Procedures for the evaluation of uncertainities
[5] Frederic Willauer 2008 Pressure Measurement (France: Endress+ Hauser)
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 052005 doi:10.1088/1755-1315/15/5/052005
8