Application of the most intelligent pipeline leaka

  • Detail

Application of intelligent pipeline leakage monitoring system in product oil pipeline

abstract according to the situation of Ke Wu line, using "HKH series pipeline leakage monitoring alarm positioning system", a suitable special fuzzy neural network is designed, and a special network diagram and some algorithms are given. The debugging method is explained, the test data is analyzed, the applicability of the system in the k-U product oil pipeline is concluded according to the test effect, and the existing problems and improvement suggestions are put forward

keywords artificial intelligence; Blur neural collaterals; Leakage monitoring; Application Research

Chinese Library Classification Number: document identification code: article number

1 introduction

artificial intelligence, as an important part of modern design methodology ○ 1, is an important part of some laboratory machine parts design manuals and laboratory machine manufacturing monographs, and has been successfully applied in pipeline leakage monitoring in China ○ 2. In order to solve the technical problems of the leakage monitoring of the product oil pipeline in Ukraine and make up for the shortcomings of foreign technology, according to the situation of the pipeline between Ukraine and Ukraine, the "HKH series pipeline leakage monitoring alarm positioning system" is adopted to study the appropriate special fuzzy neural network and determine the applicability of the network on the product oil pipeline between Ukraine and Kosovo

2 artificial intelligence pipeline leakage monitoring model

2.1 neural network model for leakage monitoring of g-wu product oil pipeline

g-wu product oil pipeline is a long-distance and multi variety oil closed sequential transmission pipeline, with two oil inlet stations, one oil outlet station, and four series pump stations in the middle. Each pump station also has frequency conversion pumps and power frequency pumps. The series pump station has no flowmeter, and the number of oil mixing heads in the pipeline is variable, Pipeline transmission pressure changes greatly and the situation is complex. In order to correctly identify the pipeline leakage signal in this environment, after repeated research, we have designed a special network. In order to provide as much information as possible for the network, we have added a differential pressure transmitter in the station without a flowmeter to replace the flowmeter signal. In this way, each station has three kinds of original signals entering the network

a B C D E F G h

the network is divided into eight layers

a layer: it is the input layer, where:

is the threshold set manually, which is a value related to the nature and range of the detected quantity

is the input quantity. For the middle station, they represent the inlet pressure, differential pressure and outlet pressure respectively

b layer is the fuzzy layer: the fuzzification of the detection quantity is completed here, which is the output of the upper layer and the input of the lower layer


where: is a coefficient related to the nature and range of the detected quantity,

is a quantity related to the acquisition number

c layer: it is the threshold configuration layer, where the threshold configuration before information classification is completed

is a manually set threshold value. For any input, each subordinate unit corresponding to this value is different


here, 1, 2, 3; 1,2,3 12。

is a signal distributor. For any input, each subordinate unit is different


is the threshold associated with

d layer is the classification layer: here, the pipeline operation state is divided into 12 feature points:

the last calculation result is saved,


e layer is the historical experience layer:

is the extracted historical experience, which is constantly updated after certain operations according to the data changes of F layer. When the data of F layer comes in, this layer should identify the update requirements, After reasonable treatment, it is preserved as a new historical experience


f layer is the function operation layer:

includes two aspects: one is the comprehensive operation of polycaprolactone (PCL) according to previous experience and current incoming information, giving an output data to the next level, and the other is giving a new data of historical experience to the E layer


g layer is the composite layer:

a new value synthesized according to the previous level data is sent to the next level


h layer is the output layer:

accept the first three data, and output it after this operation, giving a unique conclusion


2.2 positioning technology of special fuzzy neural network

dynamic pipeline leakage monitoring has only two problems to solve, one is to correctly identify the leakage, the other is to locate. Identification of leakage is the key technology, and positioning is based on it

using special fuzzy neural network monitoring can correctly identify pipeline leakage, and this first problem is solved. Next is the problem of positioning

in order to solve the problem of large positioning error under small signal of Kewu line, we choose the water hammer wave velocity method. According to the length of the pipeline in the monitoring point, the time difference between the upstream and downstream monitoring points receiving the water hammer wave signal and the propagation speed of the water hammer wave in the pipeline, the location of the leakage point can be calculated. The positioning formula of this method is as follows:



location of leakage point

length of monitored pipeline

speed of wave propagation in the pipeline

time difference between the first and last stations receiving waves



fluid density

k volumetric elastic coefficient of liquid

Realize commercialization as soon as possible

e pipe elastic coefficient

d average diameter of pipe

pipe wall thickness

coefficient, for buried pipelines, = 1 -

Poisson coefficient, of steel pipe =0.3

it can be seen from the above formula that in general, the change of sum can be ignored, and the size of positioning error is only related to. Therefore, the key problem of using water hammer wave velocity to calculate the leakage position is to find the exact time of the event. Because the special fuzzy neural network can accurately find the time of the event, it provides a basis for accurate calculation, and also solves the problem of too long positioning time for small signals

3 debugging of special fuzzy neural network

debugging of special artificial intelligence network is carried out on site after the installation of system software and hardware. First, carry out preliminary commissioning according to the operating parameters, and then carry out field oil drainage test. There is no debugging operation in the oil drainage process. After the oil drainage, the system learning model is adjusted according to the oil drainage results, which is a continuous cycle process. Until the commissioning target is reached

commissioning oil drainage method: during the period from 1 hour before oil drainage to the end of the last oil drainage, each station should not carry out major manual operations, including starting and stopping the pump, switching oil mixing and adjusting the frequency conversion equipment; Each unit of the system operates normally for more than 30 minutes; The shortest time interval between two oil discharges shall not be less than 20 minutes (from the previous valve closing time to the next valve opening time); The on-site oil drainage operation shall be accurate and fast, and the accurate oil drainage measurement shall be carried out; The minimum discharge volume shall not be less than 0.5% of the instantaneous volume; Open and close the valve to drain oil continuously, and do not open and close the valve repeatedly to adjust the flow; The duration of each oil drain shall not be less than 2.5 minutes

4 system test results

under the conditions of normal oil transportation and stable operation of the system, the downstream pipeline that is not easy to alarm is selected as the test section, and the performance and characteristics of the "HKH series pipeline leakage monitoring, alarm and positioning system" are actually tested by the method of on-site oil drainage. In the whole process, the system is in a fully automatic and unattended state

4.1 test data when transporting a single type of oil

Table 1 on October 25, 2004, oil discharge field test data table 135km away from 703 station

serial number: divided oil delivery m3/h oil discharge m3/h oil discharge speed% alarm situation position error km

1 12:48 309.1 2 0.679 134.69 -0.31

2 13:00 301.2 1 0.564 130 -4.24

3 13:31 301.2 2 2 2.0 0.664 135.9 0.9

4 14:00 300.1 1.5 0.500 134.86 -0.14

the number in Table 1 According to the field records, the oil delivered during the test was diesel, and there was no mixing head in the pipeline. The corresponding time of each oil drain alarm is less than 3 minutes

4.2. On September 25th, 2005, when there was a gasoline mixing head in the pipeline, the test was carried out again

Table 2 data sheet of on-site oil discharge test when there is a mixing head in the pipeline

oil discharge place oil discharge time September 25th, 2005 oil delivery volume m3/h oil discharge speed m3/h (oil discharge volume/oil delivery volume) 100% of the alarm position is km away from the initial station. Response time min

automatic recording of field measurement automatic recording of field measurement

10 valve pool 11:: 40 278 0.692 1.9 0.25 0.68 238.5 3

12:: 25 278 1.9 0.68

12:: 00 278 1.6 0.58

13:: 40 278 0.518 1.56 0.19 0.57 239.88 3

14:: 10 276 0.696 1.77 0.25 0.64 237.65 3

13 valve pool 16:: 30 277 0.717 1.32 0.26 0.48 260.04 3

16:: 00 277 0.609 1.4 0.2 2 0.51 272.243 3

17:: 40 277 1.35 0.49

18:: 10 277 1.35 0.48

18:: 45 277 1.4 0.51

Table 2 automatic record is the value in the system alarm record, which is the cumulative change of the difference between the flow meters at both ends of the pipeline, and it reflects the value of the detectable signal; The on-site measurement is the measured value of the on-site flowmeter during oil drainage, which is the actual value of on-site oil drainage

4.3 data analysis

the field data and the data recorded by the monitoring system are reliable. However, the leakage data recorded by the pipeline leakage monitoring system is much smaller than the data recorded on site. This difference is determined by the characteristics of oil pipelines

4.3.1 stage characteristics after leakage

when pipeline leakage occurs, the differential pressure inside and outside the pipeline means that the fluid leaks out quickly, resulting in the local pressure drop of the pipeline. The differential pressure from the leakage point to the upstream is the largest, to the downstream is the smallest, and the leakage speed is the largest. The duration of this stage is TF, and the instantaneous leakage speed of the fluid is recorded as QF

as the pipeline shrinks and the oil pressure in the pipe decreases, the volume expands, resulting in QF greater than the leakage rate QD recorded by the monitoring system. The leakage rate on site is the lowest, which is the second stage of leakage

the leakage causes the flow of the pump to increase, and the pipeline pressure rises slightly, resulting in the increase of the leakage rate. At this time, although there are still fluctuations, the pipeline pressure has basically stabilized, which is the third stage after the leakage occurs

these three stages are the leakage transition stage. First, there was leakage QF on site, and the leakage at the monitoring point was 0. Then the monitoring system began to detect the leakage speed QD, and the pipeline was in the release stage of compression energy and overcharge energy. At this stage, QF and QD exist at the same time, but there is always QF> QD

leakage redistributes the pipeline pressure, and the pipeline gradually begins to enter a new steady-state process. This process is a stable leakage stage. QF and QD gradually approach, and finally tend to be equal

as the oil drain valve is closed, the pressure at the leakage point suddenly rises, and there is no flow on site, and there will be flow in the monitoring system. QD ends only after the field flowmeter stops running for the same TF time. After that, the pipeline enters the overcharge stage of pressure rise and flow. However, neither the on-site flowmeter nor the monitoring system can record this filling volume

from the above analysis, it can be seen that before the leakage reaches a stable state, there is always QF> QD. If the valve is closed before the opening oil drain time reaches the stable stage, the result must be that the on-site oil drain volume is greater than the amount counted by the monitoring point. It can be seen that it is an inevitable law to drain oil in a short time, with more on-site leakage and less process instrument records

how long does it take for the pipeline pressure to balance? This problem is very complicated. Someone has made a calculation ③. When a 0.5% leakage occurs in a product oil pipeline, the first wave pressure drop on the monitoring point 80km away from the leakage point is 300pa, and it decreases by about 10 after 480 seconds

Copyright © 2011 JIN SHI