ML20065M530
| ML20065M530 | |
| Person / Time | |
|---|---|
| Site: | Midland |
| Issue date: | 10/12/1982 |
| From: | Stoner M STONER ASSOCIATES, INC. |
| To: | |
| Shared Package | |
| ML20065M517 | List: |
| References | |
| NUDOCS 8210210301 | |
| Download: ML20065M530 (22) | |
Text
l e
APPENDIX C NATURAL GAS PIPELINE LEAK DETECTION SYSTEM I
a 8210210301 821019 PDR ADOCK 05000329 A
~
PIO10 SED GAS PIPELINE IEAh DETECTION SYSTD1 CONSUMERS POhER COMPANY-MIDIAND POWER PIANT by Michael A. Stoner, Ih.D.
Stoner Associates, Inc.
P. O. Box 86 Carlisle, PA 17013 (717) 243-9212 i
October 12, 1982
TABLE OF 00NTENTS PAGE LIST OF FIGURES ii INTRODUCTICt; 1
GENERAL OVERVIEW 2
DESCRIPTION OF THE SYSTm 4
MCDELING PROCESS 6
MNI'IORING SYSTE 8
OPERATION OF SHUIGFF VALVE AT STEART RD.
10 REFERENCES 12 9
i
LIST OF FIGURE:S PAGE:
FIGURE 1 - SCHENATIC OF GAS DISIRIBUTION PIPELINE SERVING MIDIAND POWER PIANr 13 FIGURE: 2 - X-t DIAGRAM EDR IODEL SOIUTION-14 FIGURE 3 - MODEL CAIGIATION CNCEPr 15 FIGURE 4 - DIAGRAM SHOWING THE RELATIONSHIP OF ME:ASURED AND 00hPUIED VALUES 16 FIGURE 5 - FIOW VERSUS TIME - Q, and Qcalculated 17 FIGURE 6 - FION VERSUS TIME SHOWING INTEGRATION
~
WINDOW INIERVAIS 18 FIGURE: 7 - IEAK RATE LIMITS VERSUS WINDOW INTERVAL 19 i
l i
ii
INTRODUCTICN ne Midland Nuclear Power Plant owned by Consmers Ibwer Cornpany has five boilers that use natural gas as their fuel, hese boilers are fed frm a Consmers Power Company gas trar. mission line through the Stewart Ibad Regulator, which is approximately twa miles West of the Midland plant site. Gas is delivered to the North side of the Midland plant site through approximately 21/2 miles of six-inch and eight-inch pipeline that is operated at 350 to 400 PSIG. Se total connected load of all five boilers is approximately one million cubic feet per hour.
It has been determined that it would be undesirable to have natural gas contaminate the air supply used for various purposes at the Midland plant.
Consmers Power Cm.pany has retained Stoner Associates, Inc. to reccanend a procedure for detecting leaks and preventing the escape to the atmosphere of large quantitiies of natural gas.
i l
Re proposed leak detection procedure, outlined in the body of this report, uses l
l a cmbination of pressure and flow measurements taken on a short time interval, a digital emputer based pipeline pressure flow model, and a fail-close shutoff valve at the Stewart Ibad Regulator Station. Rese cmbine to form a high quality leak detection procedure that should be able to detect leaks in the order i
of 10 to 20 MCFH in a lapsed time of one to two minutes and close the shutoff
(
valve.
l t
i 1
~
GENERAL OVERVIEW
'Ibere are several methods of detecting leaks on pressurized piping systes, all having different methods of accuracy and detection times. Detection time as used here indicates the elapsed time between inception and detection. %ere are two techniques in use today which provide relatively short detection times. Rese are the accustic leak detection procedure, which " listens for" leaks and cmputer-based modeling approaches which simulate the pressure-flow respnse of the system, and em. pare it with actual measurements. Se method proposed is a 5
cmputer-based systs.
he key cmponents of this computer based system are as follows:
1.
Measurements of pressure, flow, and temperature (P, Q, T) taken at approximately 5 second intervals. For the concept to work, all flow into and out of the system must be measured.
2.
A model which uses some of the (P, Q, T) measured values as boundary conditions to predict values for the remaining (P, Q, T) variables which can be empared with other measurements.
l 3.
A computer which performs the task of collecting the measurements, executing the model calculations, performing cmparisons for leak criteria evaluation, and issuirg instructions to shut off the flow to l
the systs, and alarm the operator should an abnormality (leak) exist on the system.
I l
2 1
'Ihis leak detection procedure has been applied previously to a crude oil pipeline running through the envirornnentally sensitive Gulf of Mexico. he pipeline is owned by the 01evron Oil Cmpany and transports crude oil frm anpire, Mississippi to Pascagoula, Louisiana. %e distance is approximately 100 miles.
'Ihe design flow rate is 160 barrels per minute. he present two minute leak criteria is approximately five barrels, which represents a 2 percent leak. A comparison of measured flows and calculated flows for this systen can be found on page 202 of Beference 1.
3
DESCRIPTION OF 'mE SYSTEM Figure 1 shows that portion of the Constzners Ibwer Company Distribution Systen that supplies natural gas to the tw auxiliary boilers and three HP boilers at the Midland Nuclear Power Plant. Se system is approximately 2-1/2 miles in length and consists of 8-inch and 6-inch pipes. me key components, or locations, have been identified on Figure 1, and are described below.
Point A is the Stewart M. Gate Station which receives flow from the Constzners Ibwr Transmission Line and regulates the pressure to approximately 350-400 PSIG.
Bis station will be modified to include an autm.atic shut off valve that will be progravned to fail closed. Se pipeline between Stewart M. and Point B is approximately 0.80 miles of 8-inch line.
Point B is a junction where the pipeline size reduces to six inches. te flow leaves the system at this point. From Point B to Ibint C, the pipeline is 6-inches in diameter and approximately 0.79 miles in length.
Point C is the location where gas is removed from the system to supply a anall load (2 MCFH) in the outage building. From Point C to the Midland Plant boilers is amoximately 0.84 miles of six-inch line.
l Point D is a representation of the boiler loads at the Midland Plant. Bere are five separate boilers, any or all of which can be operating. Se total connected load of all five boilers is aEproximately one million cubic feet per hour. Se following measurements will be required at the locations given.
4
l ICINr A - he flow into the 8-inch line he pressure on the outlet side of the regulator he taperature on the outlet side of the regulator he open or closed status of the shutoff valve i
n ese it m s will be measured by a remote terminal unit and transmitted to a ecmputer facility in the evaporator building control rocm.
EOINT B - No measurements are required at Ibint B; however, the model will perform cenputations at this point due to the i
dimeter change.
ICINT C - Se flow constned at tihe Outage Building he pressure 2e tmperature ICINT D - Se AP across each orifice plate at each of the five boilers he taperature at each of the five orifice measurements to the boilers.
We static pressure upstrem of the orifice plate l
ne status of ignition of each of the five boilers l
Re pressure of gas in the pipeline upstrem of the regulator at the Midland Plant he taperature of the gas in the pipeline upstrem of the j
regulator 1
5
ne measurements will be made on approximately a five-second scan time.
Measurements at Point A and Ibint C will be made by Remote Terminal thits (RTU's) and transmitted over leased telep'une lines to a receiving unit at the cmputer located in the boiler building. Measurements of the inputs at Point D will be hard wired to the cmputer.
POLELING PROCESS ne flow of natural gas in a piping systen such as that described above can be mcdeled with considerable accuracy using a mathematical description of the continuity, mmentun, and energy relationships pertaining to the flow of fluids in closed conduits. Cne such model is detailed on Pages 321 - 325 of Wylie and Streeter (Eeference 1).
%ese equations have been progranmed in a general purpose model called GASTHEBM (Reference 2). A forerunner of the GASTHEBM model was used to validate the use of GASUS for the research work done by Mechanics Research, Inc. entitled " Nuclear Power Plant Risks From a Natural Gas Pipeline" as it related to the Hartsville Nuclear Ibwer Plant, Hartsville, 'IN.
GASTHEIN was developed as a general purpose model in 1979 for modeling tmsteady l
pressure flow and temperature in the Northwest Alaskan Gas Pipeline. his model has been used on various projects over the last three years. %e calculation procedure for this model results frm applying the method of characteristics to the three partial differential equations mentioned above, which produces model equations with irdependent variables, X, t (Metance along the pipeline, and time) and with dependent variables, P, Q and T (pressure, flow, and temperature).
Figure 2 shows an X-t diagram for the P, Q, T model. Assume for the mment a simplified pipeline with upstrean end at A, and dcwnstream end at D.
A typical solution of the model muld be to calculate the predicted flow at the tpstream end, Q at time t + at, knowing the pressure and temperature (P T ) at a given 6
~
time, and to calculate the downstrem pressure and ternperature (PD D} #
+0 knowing the downstrem flow, g at t.
This procedure can be repeated for time t
+ 2 at in a like manner as long as values of P, T, d g are avaMle at the g
g new time. %e procedure for using these calculated variables in a leak detection sense is described in a later section of this report.
%e items that must be assuned to provide enough information for the model to function include the gas properties, the friction parmeters for the pipeline, and an estimate of ground temperature.
It is proposed to use the pipeline calculation algorithn from GASTHEPM as the model in this leak detection process. Se equations, as they are solved in GASTHER4 usirg the method of characteristics, have been verified by cmparison with a closed form analytical model. We conparison can be seen on the attached figures, taken fran the abstract of a paper by Dr.
C. P. Liou, Ih.D., of Stoner Associates, Inc. and Professor E. B. Wylie of the Ohiversity of Michigan entitled "Che Dimensional Transient Gas Flow With Internal Heating", a copy of which is attached to this report. He emparison is performed for an internal heating process in a pipeline. The agrement between GASTHEMI and the closed form analytical solution is very good.
i An isothermal model for unsteady flow in gas systens would probably be quite adequate for the purposes intended as temperature effects in a gas distribution systen are usually of second order importance. Reference 3, page 52, shows a conparison of pressures recorded versus those calculated for a natural gas pipeline having the same approximate flowrates and pressure drops as encountered in the Midland Pipeline Systen. The proposed GASTHEPM calculation algorithn is t
expected to be more rigorous as tenperature effects are considered.
l 7
MONITORING SYSTEM The emplete monitoring systen for the proposed leak detection applir = Hon is
. shown in a simplified form in Figure 3.
Measurements will be taken at Ibints A, C, and D, on approximately a five-second interval. The model will use some of these measurements as driving conditions and will calculate the remaining variables. Specifically, the flow at Point A will be calculated and the pressure and temperature at Point D will be calculated. The flow at Point C will be specified and the temperature and pressure will be calculated.
Figure 4 shows this process in some detail and shows the relationship between the model calculation time, which will be on a fixed timestep ( a t) and the non-uniform interval between measurements brought back frm the Rru's. The a t of the calculation procedure will be either 1.6 or 3.2 seconds. The elapsed time between measurements will be on the order of five seconds. Straight line interp:)lation is used to obtain estimated values of measurements in between two actual measurement points.
Figure 5 shows the flow at Ibint A as a function of time. Indicated on the graph are both the measured flow, O and the calculated flow O As long as the g
g.
calculated and the measured values agree to within some sna11 tolerance, the integrity of the systen is assuned. This is depicted by the close cmparison of the two flows up to time, t.
If a leak or other abnormality develops on the system, the difference between measured and calculated values of flow will increase, indicating that a problem exists.
8
he primary leak detection criteria will be based on a cmparison of the measured flow and calculated flow at Mint A (Stewart M.).
Various characteristics of the metering systen affeet the Eoint emparison of calculated versus measured values in such a way as it has been beneficial to have several different cmparison criteria. Stated another way, given a level of certainty, large leaks can be detected in a short period of time, whereas smaller leaks will take longer to detect. This leads to the concept shown in Figure 6 of having different voline integration windows. For instance, th'e acetnulated flow error indicated as AFE is the stmnation of measured versus calculated flow difference times a t y
for the ntaber of calculations performed during the window interval, W. Wen a 1
new calculation is available at, for instance, t + a t, the new value is added into the stamation and th'e oldest value is dropped. For the proposed systen, four windows would be used, noteably a single point in time emparison (zero window length), a thirty-second window (about 6 measured point emparisons), a te-minute window, and a 10-minute window. An instantaneous comparison would also be made between each of the calculated and measured pressures at Point C and l
Point D.
As mentioned above, smaller leaks take a longer time to detect. This is shown graphically in Figure 7 which shows the relationship between the acetnulated flow error expressed as an average flow rate, and the window interval time. It is expected that the real leak limits can be on the order of tw percent of maxista throughput flow (aEproximately 20 MCFH) for the 30 second window and one percent (10 MCFH) for the two-minute window. This translates into leak volume limits of approximately 150 SCF for 30 seconds and 300 SCF for two minutes leak duration.
I l
l 9
l i
~
Figure 7 also shows the use of the pseudo leak limit. Se monitoring system is designed such that, as operating experience is gained, the leak limits can be adjusted. %e pseudo limits provide this capability in that they allow a tighter leak detection limit that generates leak reports internally for a systems analyst, but does not alarm the operator. He systs will save the last 30 minutes of operation information thereby allowing an analyst to tme the systen, and investigate any problens that arise in the operation of the system. As the system is tmed, and the tolerances tightened, the real leak limits should be set such that they generate no more than approximately one false alarm per year.
When the system determines that the real leak limit has been exceeded, it will initiate the closing of the shutoff valve at Stewart M. and will notify the operator.
%e systen will also provide operator display and test functions.
OPERATICN OF SHUIOFF VALVE AT STENGT RD.
Since the shutoff valve at Stewart M. is the major device to deny flow to a potential line break, it is proposed that this valve have a fail-closed mode of operation.
It should operate in such a way that it must receive a positive i
signal from the monitorirg system, to open or to stay open. %e conditions under which it would closo muld be the following:
1.
An interruption in electric power would fail to hold it open, and it muld close either under the force of gravity or under pneunatic power.
10
2.
If the signal holding it open is interrupted it will initiate closure.
3.
If the valve does not receive a positive signal from the computer based monitoring system every ten seconds, it will initiate closure. Ebr the monitoring systs to function, it will have to be receiving data frm the RTUs on every scan which means that the data acquisition systen will have to be operating properly for a positive signal to be sent to the valve.
4.
It will initiate closure umediately if it receives a signal frm the monitoring system to close. 'Ihe monitoring systen will be programed such that the leak detection procedure must be operating for it to send a psitive signal to the valve which will allow the shutoff valve to open.
i l
1 l
l
[
l l
11
REFERENCES 1.
Wylie, E. Benjmin and Streeter, V. L: Fluid Transients, McGraw Hill, Inc.,
1978.
2.
GASitiEIE Service User's GLlide, published by Stoner Associates, Inc.,
Carlisle, Pennsylvania, May, 1981.
3.
Stoner, M. A. : " Analysis and Control Of thsteady Flows In Natural Gas Piping Syst es", R1.D.
'Ihesis (Civil Engineering), thiversity of Michigan,- Ann Arbor, 1968.
i f
12
C 6"
84 6"
19 "
~
5 Boilers - Midland Plant Site 8".8 Mi.
A Stewart Ed. City Gate Regulation frczn Transmissian Line FIGURE 1 SCfDETIC OF GAS DISTRIBUTION PIPELME SERVING MIDIAND POWER PIANT t
13
I P,T P,T g
3 Dc gc Q
c+
Ac c
a D
At t
C t
A D
X FIGURE 2 X-t DIAGRAM FOR PODEL SOIUTION f
14
A B
C D
A A
V (P
0 T )g
-(P O
T)
(P o
T )g 3
3 3
C C
C g
g g
RTU y
t MEASUREMENTS 4 e r%
m MODEL U5ES 4
@C T }C AC C
FIRST SU2 SCRIPT - LOCATION A, B, C, D SECOND SUBSCRIPT - M: MEASURED C: CALCULATED l
l l
l l
FIGURE 3 MODEL CALCULATION CONCEPT
(P,0,T)g o u
t + 4 t, t Q
x t+3at AC x
(P,0,T)g o o
t+2at Q
x (P,0,T)g o o
t+
at Q
x O
O 0
l X (distance)
FIGURE 4 DIAGRAM SIDi1NG TDE REIATIONSHIP l
OF EASURED AND CDMPUIED VAUJES l
l l
16 l
u tA v __
- n
.'X,-
' 'x, AM g
'% X, x
Discrete measurements Q
AC x Model calculation values e-at -+
time FIGJC 5
~ b asured Ocalculated 17
a 4
O g
O e.
yt+At
=
t
,d l
t W2 a,
w Q, 3, "
,--x--....,____x----..
g g
AC ht G
I T
I t
e
~~
t-Wl
@g - Og) b
.MnR4 FILM ERPOR = AFE
=
y y
t t-W2 (O
Og) x at ACCU 4 FI4W ERROR
- AE 2
2, g
t l
FIGURE 6 l
l FILM VERSUS TDE SIOWING INIEGRTION wmxw nnERvAIs 18
(
~o-n-a.s-u
---a y-m.
-+
4 4
e 9
N Real Ieak Limit
- Pseudo Ieak Limit g
WI e
9 e
W W
W l
2 3
1 1
I J
WINDOW INTERVAL OfI)
(IEAK DETECTION TIME)
FIGURE 7 LEAK RATE LIMITS VERSUS WINDOW INIERVAL 4
4 19
- w-
--g.---w--
y
+my-i- -,----, --
,ep.4 y----
-wwgy.g---.
y, --,,,,,
y, y
p y---
.------- - -